hwaccel docs

add support for iHD driver
ensure dummy frame is in yuv shape
2025-11-03 01:43:35 +08:00 · 2020-10-25 14:30:36 -05:00 · 2020-10-25 14:30:36 -05:00 · 2020-10-25 14:30:36 -05:00 · 2020-10-25 14:30:36 -05:00 · 2020-10-25 14:30:36 -05:00
29 changed files with 2941 additions and 912 deletions
--- a/.dockerignore
+++ b/.dockerignore
@@ -1,5 +1,5 @@
 README.md
-diagram.png
+docs/
 .gitignore
 debug
 config/
--- a/63
+++ b/63
@@ -1,63 +0,0 @@
-FROM ubuntu:18.04
-LABEL maintainer "blakeb@blakeshome.com"
-
-ENV DEBIAN_FRONTEND=noninteractive
-# Install packages for apt repo
-RUN apt -qq update && apt -qq install --no-install-recommends -y \
-    software-properties-common \
-    # apt-transport-https ca-certificates \
-    build-essential \
-    gnupg wget unzip tzdata \
-    # libcap-dev \
-    && add-apt-repository ppa:deadsnakes/ppa -y \
-    && apt -qq install --no-install-recommends -y \
-        python3.7 \
-        python3.7-dev \
-        python3-pip \
-        ffmpeg \
-        # VAAPI drivers for Intel hardware accel
-        libva-drm2 libva2 i965-va-driver vainfo \
-    && python3.7 -m pip install -U pip \
-    && python3.7 -m pip install -U wheel setuptools \
-    && python3.7 -m pip install -U \
-        opencv-python-headless \
-        # python-prctl \
-        numpy \
-        imutils \
-        scipy \
-        psutil \
-    && python3.7 -m pip install -U \
-        Flask \
-        paho-mqtt \
-        PyYAML \
-        matplotlib \
-        pyarrow \
-        click \
-    && echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" > /etc/apt/sources.list.d/coral-edgetpu.list \
-    && wget -q -O - https://packages.cloud.google.com/apt/doc/apt-key.gpg | apt-key add - \
-    && apt -qq update \
-    && echo "libedgetpu1-max libedgetpu/accepted-eula boolean true" | debconf-set-selections \
-    && apt -qq install --no-install-recommends -y \
-        libedgetpu1-max \
-    ## Tensorflow lite (python 3.7 only)
-    && wget -q https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp37-cp37m-linux_x86_64.whl \
-    && python3.7 -m pip install tflite_runtime-2.1.0.post1-cp37-cp37m-linux_x86_64.whl \
-    && rm tflite_runtime-2.1.0.post1-cp37-cp37m-linux_x86_64.whl \
-    && rm -rf /var/lib/apt/lists/* \
-    && (apt-get autoremove -y; apt-get autoclean -y)
-
-# get model and labels
-RUN wget -q https://github.com/google-coral/edgetpu/raw/master/test_data/ssd_mobilenet_v2_coco_quant_postprocess_edgetpu.tflite -O /edgetpu_model.tflite --trust-server-names
-COPY labelmap.txt /labelmap.txt
-RUN wget -q https://github.com/google-coral/edgetpu/raw/master/test_data/ssd_mobilenet_v2_coco_quant_postprocess.tflite -O /cpu_model.tflite 
-
-
-RUN mkdir /cache /clips
-
-WORKDIR /opt/frigate/
-ADD frigate frigate/
-COPY detect_objects.py .
-COPY benchmark.py .
-COPY process_clip.py .
-
-CMD ["python3.7", "-u", "detect_objects.py"]
--- a/37
+++ b/37
@@ -0,0 +1,37 @@
+default_target: amd64_frigate
+
+amd64_wheels:
+	docker build --tag blakeblackshear/frigate-wheels:amd64 --file docker/Dockerfile.wheels .
+
+amd64_ffmpeg:
+	docker build --tag blakeblackshear/frigate-ffmpeg:amd64 --file docker/Dockerfile.ffmpeg.amd64 .
+
+amd64_frigate:
+	docker build --tag frigate-base --build-arg ARCH=amd64 --file docker/Dockerfile.base .
+	docker build --tag frigate --file docker/Dockerfile.amd64 .
+
+amd64_all: amd64_wheels amd64_ffmpeg amd64_frigate
+
+aarch64_wheels:
+	docker build --tag blakeblackshear/frigate-wheels:aarch64 --file docker/Dockerfile.wheels.aarch64 .
+
+aarch64_ffmpeg:
+	docker build --tag blakeblackshear/frigate-ffmpeg:aarch64 --file docker/Dockerfile.ffmpeg.aarch64 .
+
+aarch64_frigate:
+	docker build --tag frigate-base --build-arg ARCH=aarch64 --file docker/Dockerfile.base .
+	docker build --tag frigate --file docker/Dockerfile.aarch64 .
+
+armv7_all: armv7_wheels armv7_ffmpeg armv7_frigate
+
+armv7_wheels:
+	docker build --tag blakeblackshear/frigate-wheels:armv7 --file docker/Dockerfile.wheels .
+
+armv7_ffmpeg:
+	docker build --tag blakeblackshear/frigate-ffmpeg:armv7 --file docker/Dockerfile.ffmpeg.armv7 .
+
+armv7_frigate:
+	docker build --tag frigate-base --build-arg ARCH=armv7 --file docker/Dockerfile.base .
+	docker build --tag frigate --file docker/Dockerfile.armv7 .
+
+armv7_all: armv7_wheels armv7_ffmpeg armv7_frigate
--- a/README.md
+++ b/README.md
@@ -1,73 +1,513 @@
+<p align="center">
+  <img width="40%" align="center" alt="logo" src="docs/frigate.png">
+</p>
+
 # Frigate - NVR With Realtime Object Detection for IP Cameras
 Uses OpenCV and Tensorflow to perform realtime object detection locally for IP cameras. Designed for integration with HomeAssistant or others via MQTT.

-Use of a [Google Coral Accelerator](https://coral.ai/products/) is optional, but highly recommended. On my Intel i7 processor, I can process 2-3 FPS with the CPU. The Coral can process 100+ FPS with very low CPU load.
+Use of a [Google Coral Accelerator](https://coral.ai/products/) is optional, but highly recommended. The Coral will outperform even the best CPUs and can process 100+ FPS with very little overhead.

+- Designed to minimize resource use and maximize performance by only looking for objects when and where it is necessary
 - Leverages multiprocessing heavily with an emphasis on realtime over processing every frame
 - Uses a very low overhead motion detection to determine where to run object detection
- Object detection with Tensorflow runs in a separate process
- Object info is published over MQTT for integration into HomeAssistant as a binary sensor
- An endpoint is available to view an MJPEG stream for debugging, but should not be used continuously
+- Object detection with TensorFlow runs in separate processes for maximum FPS
+- Communicates over MQTT for easy integration into other systems

-![Diagram](diagram.png)
+## Documentation
+- [How Frigate Works](docs/how-frigate-works.md)
+- [Recommended Hardware](#recommended-hardware)
+- [Installing](#installing)
+- [Configuration File](#configuration)
+- [Optimizing Performance](#optimizing-performance)
+- [Detectors](#detectors)
+- [Object Filters](#object-filters)
+- [Masks](#masks)
+- [Zones](#zones)
+- [Integration with HomeAssistant](#integration-with-homeassistant)
+- [MQTT Topics](#mqtt-topics)
+- [HTTP Endpoints](#http-endpoints)
+- [Custom Models](#custom-models)
+- [Troubleshooting](#troubleshooting)

-## Example video (from older version)
-You see multiple bounding boxes because it draws bounding boxes from all frames in the past 1 second where a person was detected. Not all of the bounding boxes were from the current frame.
-[![](http://img.youtube.com/vi/nqHbCtyo4dY/0.jpg)](http://www.youtube.com/watch?v=nqHbCtyo4dY "Frigate")
+## Recommended Hardware
+|Name|Inference Speed|Notes|
+|----|---------------|-----|
+|Atomic Pi|16ms|Good option for a dedicated low power board with a small number of cameras. Can leverage Intel QuickSync for stream decoding.|
+|Intel NUC NUC7i3BNK|8-10ms|Great performance. Can handle many cameras at 5fps depending on typical amounts of motion.|
+|BMAX B2 Plus|10-12ms|Good balance of performance and cost. Also capable of running many other services at the same time as frigate.|
+|Minisforum GK41|9-10ms|Great alternative to a NUC with dual Gigabit NICs. Easily handles several 1080p cameras.|
+|Raspberry Pi 3B (32bit)|60ms|Can handle a small number of cameras, but the detection speeds are slow due to USB 2.0.|
+|Raspberry Pi 4 (32bit)|15-20ms|Can handle a small number of cameras. The 2GB version runs fine.|
+|Raspberry Pi 4 (64bit)|10-15ms|Can handle a small number of cameras. The 2GB version runs fine.|

-## Getting Started
-Run the container with
-```bash
-docker run --rm \
-name blakeblackshear/frigate:stable \
--privileged \
--shm-size=512m \ # should work for a 2-3 cameras
-v /dev/bus/usb:/dev/bus/usb \
-v <path_to_config_dir>:/config:ro \
-v /etc/localtime:/etc/localtime:ro \
-p 5000:5000 \
-e FRIGATE_RTSP_PASSWORD='password' \
-blakeblackshear/frigate:stable
-```
+[Back to top](#documentation)

-Example docker-compose:
+## Installing
+
+### HassOS Addon
+HassOS users can install via the addon repository. Frigate requires that an MQTT server be running.
+1. Navigate to Supervisor > Add-on Store > Repositories
+1. Add https://github.com/blakeblackshear/frigate-hass-addons
+1. Setup your configuration in the `Configuration` tab
+1. Start the addon container
+
+### Docker
+Make sure you choose the right image for your architecture:
+|Arch|Image Name|
+|-|-|
+|amd64|blakeblackshear/frigate:stable-amd64|
+|armv7|blakeblackshear/frigate:stable-armv7|
+|aarch64|blakeblackshear/frigate:stable-aarch64|
+
+It is recommended to run with docker-compose:
 ```yaml
  frigate:
    container_name: frigate
    restart: unless-stopped
    privileged: true
-    shm_size: '1g' # should work for 5-7 cameras
-    image: blakeblackshear/frigate:stable
+    image: blakeblackshear/frigate:stable-amd64
    volumes:
      - /dev/bus/usb:/dev/bus/usb
      - /etc/localtime:/etc/localtime:ro
      - <path_to_config>:/config
-      - <path_to_directory_for_clips>:/clips
+      - <path_to_directory_for_clips>:/clips      
+      - type: tmpfs # 1GB of memory, reduces SSD/SD Card wear
+        target: /cache
+        tmpfs:
+          size: 100000000
    ports:
      - "5000:5000"
    environment:
      FRIGATE_RTSP_PASSWORD: "password"
+    healthcheck:
+      test: ["CMD", "wget" , "-q", "-O-", "http://localhost:5000"]
+      interval: 30s
+      timeout: 10s
+      retries: 5
+      start_period: 3m
 ```

-A `config.yml` file must exist in the `config` directory. See example [here](config/config.example.yml) and device specific info can be found [here](docs/DEVICES.md).
+If you can't use docker compose, you can run the container with:
+```bash
+docker run --rm \
+--name frigate \
+--privileged \
+-v /dev/bus/usb:/dev/bus/usb \
+-v <path_to_config_dir>:/config:ro \
+-v /etc/localtime:/etc/localtime:ro \
+-p 5000:5000 \
+-e FRIGATE_RTSP_PASSWORD='password' \
+blakeblackshear/frigate:stable-amd64
+```

-## Recommended Hardware
-|Name|Inference Speed|Notes|
-|----|---------------|-----|
-|Atomic Pi|16ms|Best option for a dedicated low power board with a small number of cameras.|
-|Intel NUC NUC7i3BNK|8-10ms|Best possible performance. Can handle 7+ cameras at 5fps depending on typical amounts of motion.|
-|BMAX B2 Plus|10-12ms|Good balance of performance and cost. Also capable of running many other services at the same time as frigate.|
-|Minisforum GK41|9-10ms|Great alternative to a NUC. Easily handiles 4 1080p cameras.|
+### Kubernetes
+Use the [helm chart](https://github.com/k8s-at-home/charts/tree/master/charts/frigate).

-ARM boards are not officially supported at the moment due to some python dependencies that require modification to work on ARM devices. The Raspberry Pi4 gets about 16ms inference speeds, but the hardware acceleration for ffmpeg does not work for converting yuv420 to rgb24. The Atomic Pi is x86 and much more efficient.
+### Virtualization
+For ideal performance, Frigate needs access to underlying hardware for the Coral and GPU devices for ffmpeg decoding. Running Frigate in a VM on top of Proxmox, ESXi, Virtualbox, etc. is not recommended. The virtualization layer typically introduces a sizable amount of overhead for communication with Coral devices.

-Users have reported varying success in getting frigate to run in a VM. In some cases, the virtualization layer introduces a significant delay in communication with the Coral. If running virtualized in Proxmox, pass the USB card/interface to the virtual machine not the USB ID for faster inference speed.
+#### Proxmox
+Some people have had success running Frigate in LXC directly with the following config:
+```
+arch: amd64
+cores: 2
+features: nesting=1
+hostname: FrigateLXC
+memory: 4096
+net0: name=eth0,bridge=vmbr0,firewall=1,hwaddr=2E:76:AE:5A:58:48,ip=dhcp,ip6=auto,type=veth
+ostype: debian
+rootfs: local-lvm:vm-115-disk-0,size=12G
+swap: 512
+lxc.cgroup.devices.allow: c 189:385 rwm
+lxc.mount.entry: /dev/dri/renderD128 dev/dri/renderD128 none bind,optional,create=file
+lxc.mount.entry: /dev/bus/usb/004/002 dev/bus/usb/004/002 none bind,optional,create=file
+lxc.apparmor.profile: unconfined
+lxc.cgroup.devices.allow: a
+lxc.cap.drop:
+```
+
+### Calculating shm-size
+The default shm-size of 64m is fine for setups with 3 or less 1080p cameras. If frigate is exiting with "Bus error" messages, it could be because you have too many high resolution cameras and you need to specify a higher shm size.
+
+You can calculate the necessary shm-size for each camera with the following formula:
+```
+(width * height * 1.5 * 7 + 270480)/1048576 = <shm size in mb>
+```
+[Back to top](#documentation)
+
+## Configuration
+HassOS users can manage their configuration directly in the addon Configuration tab. For other installations, the default location for the config file is `/config/config.yml`. This can be overridden with the `CONFIG_FILE` environment variable. Camera specific ffmpeg parameters are documented [here](docs/cameras.md).
+
+```yaml
+# Optional: port for http server (default: shown below)
+web_port: 5000
+
+# Optional: detectors configuration
+# USB Coral devices will be auto detected with CPU fallback
+detectors:
+  # Required: name of the detector
+  coral:
+    # Required: type of the detector
+    # Valid values are 'edgetpu' (requires device property below) and 'cpu'.
+    type: edgetpu
+    # Optional: device name as defined here: https://coral.ai/docs/edgetpu/multiple-edgetpu/#using-the-tensorflow-lite-python-api
+    device: usb
+
+# Required: mqtt configuration
+mqtt:
+  # Required: host name
+  host: mqtt.server.com
+  # Optional: port (default: shown below)
+  port: 1883
+  # Optional: topic prefix (default: shown below)
+  # WARNING: must be unique if you are running multiple instances
+  topic_prefix: frigate
+  # Optional: client id (default: shown below)
+  # WARNING: must be unique if you are running multiple instances
+  client_id: frigate
+  # Optional: user
+  user: mqtt_user
+  # Optional: password
+  # NOTE: Environment variables that begin with 'FRIGATE_' may be referenced in {}. 
+  #       eg. password: '{FRIGATE_MQTT_PASSWORD}'
+  password: password
+
+# Optional: Global configuration for saving clips
+save_clips:
+  # Optional: Maximum length of time to retain video during long events. (default: shown below)
+  # NOTE: If an object is being tracked for longer than this amount of time, the cache
+  #       will begin to expire and the resulting clip will be the last x seconds of the event.
+  max_seconds: 300
+  # Optional: Location to save event clips. (default: shown below)
+  clips_dir: /clips
+  # Optional: Location to save cache files for creating clips. (default: shown below)
+  # NOTE: To reduce wear on SSDs and SD cards, use a tmpfs volume.
+  cache_dir: /cache
+
+# Optional: Global ffmpeg args
+# "ffmpeg" + global_args + input_args + "-i" + input + output_args
+ffmpeg:
+  # Optional: global ffmpeg args (default: shown below)
+  global_args:
+    - -hide_banner
+    - -loglevel
+    - panic
+  # Optional: global hwaccel args (default: shown below)
+  # NOTE: See hardware acceleration docs for your specific device
+  hwaccel_args: []
+  # Optional: global input args (default: shown below)
+  input_args:
+    - -avoid_negative_ts
+    - make_zero
+    - -fflags
+    - nobuffer
+    - -flags
+    - low_delay
+    - -strict
+    - experimental
+    - -fflags
+    - +genpts+discardcorrupt
+    - -rtsp_transport
+    - tcp
+    - -stimeout
+    - '5000000'
+    - -use_wallclock_as_timestamps
+    - '1'
+  # Optional: global output args (default: shown below)
+  output_args:
+    - -f
+    - rawvideo
+    - -pix_fmt
+    - yuv420p
+
+# Optional: Global object filters for all cameras.
+# NOTE: can be overridden at the camera level
+objects:
+  # Optional: list of objects to track from labelmap.txt (default: shown below)
+  track:
+    - person
+  # Optional: filters to reduce false positives for specific object types
+  filters:
+    person:
+      # Optional: minimum width*height of the bounding box for the detected object (default: 0)
+      min_area: 5000
+      # Optional: maximum width*height of the bounding box for the detected object (default: max_int)
+      max_area: 100000
+      # Optional: minimum score for the object to initiate tracking (default: shown below)
+      min_score: 0.5
+      # Optional: minimum decimal percentage for tracked object's computed score to be considered a true positive (default: shown below)
+      threshold: 0.85
+
+# Required: configuration section for cameras
+cameras:
+  # Required: name of the camera
+  back:
+    # Required: ffmpeg settings for the camera
+    ffmpeg:
+      # Required: Source passed to ffmpeg after the -i parameter.
+      # NOTE: Environment variables that begin with 'FRIGATE_' may be referenced in {}
+      input: rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@10.0.10.10:554/cam/realmonitor?channel=1&subtype=2
+      # Optional: camera specific global args (default: inherit)
+      global_args:
+      # Optional: camera specific hwaccel args (default: inherit)
+      hwaccel_args:
+      # Optional: camera specific input args (default: inherit)
+      input_args:
+      # Optional: camera specific output args (default: inherit)
+      output_args:
+    
+    # Optional: height of the frame
+    # NOTE: Recommended to set this value, but frigate will attempt to autodetect.
+    height: 720
+    # Optional: width of the frame
+    # NOTE: Recommended to set this value, but frigate will attempt to autodetect.
+    width: 1280
+    # Optional: desired fps for your camera
+    # NOTE: Recommended value of 5. Ideally, try and reduce your FPS on the camera.
+    #       Frigate will attempt to autodetect if not specified.
+    fps: 5
+
+    # Optional: motion mask
+    # NOTE: see docs for more detailed info on creating masks
+    mask: poly,0,900,1080,900,1080,1920,0,1920
+
+    # Optional: timeout for highest scoring image before allowing it
+    # to be replaced by a newer image. (default: shown below)
+    best_image_timeout: 60
+
+    # Optional: camera specific mqtt settings
+    mqtt:
+      # Optional: crop the camera frame to the detection region of the object (default: False)
+      crop_to_region: True
+      # Optional: resize the image before publishing over mqtt
+      snapshot_height: 300
+
+    # Optional: zones for this camera
+    zones:
+      # Required: name of the zone
+      # NOTE: This must be different than any camera names, but can match with another zone on another
+      #       camera.
+      front_steps:
+        # Required: List of x,y coordinates to define the polygon of the zone.
+        # NOTE: Coordinates can be generated at https://www.image-map.net/
+        coordinates: 545,1077,747,939,788,805
+        # Optional: Zone level object filters.
+        # NOTE: The global and camera filters are applied upstream.
+        filters:
+          person:
+            min_area: 5000
+            max_area: 100000
+            threshold: 0.8
+
+    # Optional: save clips configuration
+    # NOTE: This feature does not work if you have added "-vsync drop" in your input params. 
+    #       This will only work for camera feeds that can be copied into the mp4 container format without
+    #       encoding such as h264. It may not work for some types of streams.
+    save_clips:
+      # Required: enables clips for the camera (default: shown below)
+      enabled: False
+      # Optional: Number of seconds before the event to include in the clips (default: shown below)
+      pre_capture: 30
+      # Optional: Objects to save clips for. (default: all tracked objects)
+      objects:
+        - person      
+
+    # Optional: Configuration for the snapshots in the debug view and mqtt
+    snapshots:
+      # Optional: print a timestamp on the snapshots (default: shown below)
+      show_timestamp: True
+      # Optional: draw zones on the debug mjpeg feed (default: shown below)
+      draw_zones: False
+      # Optional: draw bounding boxes on the mqtt snapshots (default: shown below)
+      draw_bounding_boxes: True
+
+    # Optional: Camera level object filters config. If defined, this is used instead of the global config.
+    objects:
+      track:
+        - person
+        - car
+      filters:
+        person:
+          min_area: 5000
+          max_area: 100000
+          min_score: 0.5
+          threshold: 0.85
+```
+[Back to top](#documentation)
+
+## Optimizing Performance
+- **Google Coral**: It is strongly recommended to use a Google Coral, but Frigate will fall back to CPU in the event one is not found. Offloading TensorFlow to the Google Coral is an order of magnitude faster and will reduce your CPU load dramatically. A $60 device will outperform $2000 CPU.
+- **Resolution**: Choose a camera resolution where the smallest object you want to detect barely fits inside a 300x300px square. The model used by Frigate is trained on 300x300px images, so you will get worse performance and no improvement in accuracy by using a larger resolution since Frigate resizes the area where it is looking for objects to 300x300 anyway.
+- **FPS**: 5 frames per second should be adequate. Higher frame rates will require more CPU usage without improving detections or accuracy. Reducing the frame rate on your camera will have the greatest improvement on system resources.
+- **Hardware Acceleration**: Make sure you configure the `hwaccel_args` for your hardware. They provide a significant reduction in CPU usage if they are available.
+- **Masks**: Masks can be used to ignore motion and reduce your idle CPU load. If you have areas with regular motion such as timestamps or trees blowing in the wind, frigate will constantly try to determine if that motion is from a person or other object you are tracking. Those detections not only increase your average CPU usage, but also clog the pipeline for detecting objects elsewhere. If you are experiencing high values for `detection_fps` when no objects of interest are in the cameras, you should use masks to tell frigate to ignore movement from trees, bushes, timestamps, or any part of the image where detections should not be wasted looking for objects.
+
+### FFmpeg Hardware Acceleration
+Frigate works on Raspberry Pi 3b/4 and x86 machines. It is recommended to update your configuration to enable hardware accelerated decoding in ffmpeg. Depending on your system, these parameters may not be compatible.
+
+Raspberry Pi 3/4 (32-bit OS):
+```yaml
+ffmpeg:
+  hwaccel_args:
+    - -c:v
+    - h264_mmal
+```
+
+Raspberry Pi 3/4 (64-bit OS)
+```yaml
+ffmpeg:
+  hwaccel_args:
+    - -c:v
+    - h264_v4l2m2m
+```
+
+Intel-based CPUs (<10th Generation) via Quicksync (https://trac.ffmpeg.org/wiki/Hardware/QuickSync)
+```yaml
+ffmpeg:
+  hwaccel_args:
+    - -hwaccel
+    - vaapi
+    - -hwaccel_device
+    - /dev/dri/renderD128
+    - -hwaccel_output_format
+    - yuv420p
+```
+
+Intel-based CPUs (>=10th Generation) via Quicksync (https://trac.ffmpeg.org/wiki/Hardware/QuickSync)
+**Note:** You also need to set `LIBVA_DRIVER_NAME=iHD` as an environment variable on the container.
+```yaml
+ffmpeg:
+  hwaccel_args:
+    - -hwaccel
+    - vaapi
+    - -hwaccel_device
+    - /dev/dri/renderD128
+```
+
+[Back to top](#documentation)
+
+## Detectors
+By default Frigate will look for a USB Coral device and fall back to the CPU if it cannot be found. If you have PCI or multiple Coral devices, you need to configure your detector devices in the config file. When using multiple detectors, they run in dedicated processes, but pull from a common queue of requested detections across all cameras.
+
+Frigate supports `edgetpu` and `cpu` as detector types. The device value should be specified according to the [Documentation for the TensorFlow Lite Python API](https://coral.ai/docs/edgetpu/multiple-edgetpu/#using-the-tensorflow-lite-python-api).
+
+Single USB Coral:
+```yaml
+detectors:
+  coral:
+    type: edgetpu
+    device: usb
+```
+
+Multiple USB Corals:
+```yaml
+detectors:
+  coral1:
+    type: edgetpu
+    device: usb:0
+  coral2:
+    type: edgetpu
+    device: usb:1
+```
+
+Mixing Corals:
+```yaml
+detectors:
+  coral_usb:
+    type: edgetpu
+    device: usb
+  coral_pci:
+    type: edgetpu
+    device: pci
+```
+
+CPU Detectors (not recommended):
+```yaml
+detectors:
+  cpu1:
+    type: cpu
+  cpu2:
+    type: cpu
+```
+[Back to top](#documentation)
+
+## Reducing False Positives
+Tune your object filters to adjust false positives: `min_area`, `max_area`, `min_score`, `threshold`.
+
+For object filters in your configuration, any single detection below `min_score` will be ignored as a false positive. `threshold` is based on the median of the history of scores (padded to 3 values) for a tracked object. Consider the following frames when `min_score` is set to 0.6 and threshold is set to 0.85:
+
+| Frame | Current Score | Score History | Computed Score | Detected Object |
+| --- | --- | --- | --- | --- |
+| 1 | 0.7 | 0.0, 0, 0.7 | 0.0 | No
+| 2 | 0.55 | 0.0, 0.7, 0.0 | 0.0 | No
+| 3 | 0.85 | 0.7, 0.0, 0.85 | 0.7 | No
+| 4 | 0.90 | 0.7, 0.85, 0.95, 0.90 | 0.875 | Yes
+| 5 | 0.88 | 0.7, 0.85, 0.95, 0.90, 0.88 | 0.88 | Yes
+| 6 | 0.95 | 0.7, 0.85, 0.95, 0.90, 0.88, 0.95 | 0.89 | Yes
+
+In frame 2, the score is below the `min_score` value, so frigate ignores it and it becomes a 0.0. The computed score is the median of the score history (padding to at least 3 values), and only when that computed score crosses the `threshold` is the object marked as a true positive. That happens in frame 4 in the example.
+
+[Back to top](#documentation)
+
+## Masks
+The following types of masks are supported:
+- `poly`: (Recommended) List of x,y points like zone configuration
+- `base64`: Base64 encoded image file
+- `image`: Image file in the `/config` directory
+
+`base64` and `image` masks must be the same aspect ratio and resolution as your camera. 
+
+The mask in the second image would limit motion detection on this camera to only the front yard and not the street. 
+
+<a href="docs/example-mask-check-point.png"><img src="docs/example-mask-check-point.png" height="300"></a>
+<a href="docs/example-mask.bmp"><img src="docs/example-mask.bmp" height="300"></a>
+<a href="docs/example-mask-overlay.png"><img src="docs/example-mask-overlay.png" height="300"></a>
+
+To create a poly mask:
+1. Download a camera snapshot image with the same resolution as the camera feed (`/<camera_name>/latest.jpg`).
+1. Upload the image to https://www.image-map.net/
+1. Select "shape" poly - start in the lowest left corner and place the first marker (point) and continue upwards and then to the right until the polygon shape covers the area that you want to mask out (ignore). 
+1. When you are finished with the polygon click "Show me the code!" and copy all coordinates (point), ie. `"0,461,3,0,1919,0,1919,843,1699,492,1344,458,1346,336,973,317,869,375,866,432"` 
+1. Adjust any -1 values to 0 and then add it all to the configuration (see the example configuration for correct indentation and placement)
+
+Example of a finished row corresponding to the below example image:
+```yaml
+mask: 'poly,0,461,3,0,1919,0,1919,843,1699,492,1344,458,1346,336,973,317,869,375,866,432'
+```
+    
+<a href="docs/example-mask-poly.png"><img src="docs/example-mask-poly.png" height="300"></a>
+
+You can test your mask by temporarily configuring it as a [zone](#zones) and enabling`draw_zones` in your config.
+
+[Back to top](#documentation)
+
+## Zones
+Zones allow you to define a specific area of the frame and apply additional filters for object types so you can determine whether or not an object is within a particular area. Zones cannot have the same name as a camera. If desired, a single zone can include multiple cameras if you have multiple cameras covering the same area by configuring zones with the same name for each camera.
+
+During testing, `draw_zones` should be set in the config to draw the zone on the frames so you can adjust as needed. The zone line will increase in thickness when any object enters the zone.
+
+![Zone Example](docs/zone_example.jpg)
+
+[Back to top](#documentation)
+
+## Recording Clips
+**Note**: Previous versions of frigate included `-vsync drop` in input parameters. This is not compatible with FFmpeg's segment feature and must be removed from your input parameters if you have overrides set.
+
+Frigate can save video clips without any CPU overhead for encoding by simply copying the stream directly with FFmpeg. It leverages FFmpeg's segment functionality to maintain a cache of video for each camera. The cache files are written to disk at `cache_dir` and do not introduce memory overhead. When an object is being tracked, it will extend the cache to ensure it can assemble a clip when the event ends. Once the event ends, it again uses FFmpeg to assemble a clip by combining the video clips without any encoding by the CPU. Assembled clips are are saved to the `clips_dir` directory along with a json file containing the current information about the tracked object.
+
+### Global Configuration Options
+- `max_seconds`: This limits the size of the cache when an object is being tracked. If an object is stationary and being tracked for a long time, the cache files will expire and this value will be the maximum clip length for the *end* of the event. For example, if this is set to 300 seconds and an object is being tracked for 600 seconds, the clip will end up being the last 300 seconds. Defaults to 300 seconds.
+
+### Per-camera Configuration Options
+- `pre_capture`: Defines how much time should be included in the clip prior to the beginning of the event. Defaults to 30 seconds.
+- `objects`: List of object types to save clips for. Object types here must be listed for tracking at the camera or global configuration. Defaults to all tracked objects.
+
+[Back to top](#documentation)

 ## Integration with HomeAssistant
-
 Setup a camera, binary_sensor, sensor and optionally automation as shown for each camera you define in frigate. Replace <camera_name> with the camera name as defined in the frigate `config.yml` (The `frigate_coral_fps` and `frigate_coral_inference` sensors only need to be defined once)

-```
+```yaml
 camera:
  - name: <camera_name> Last Person
    platform: mqtt
@@ -90,12 +530,13 @@ sensor:
    scan_interval: 5
    json_attributes:
      - <camera_name>
-      - coral
+      - detection_fps
+      - detectors
    value_template: 'OK'  
  - platform: template
    sensors:
      <camera_name>_fps: 
-        value_template: '{{ states.sensor.frigate_debug.attributes["<camera_name>"]["fps"] }}'
+        value_template: '{{ states.sensor.frigate_debug.attributes["<camera_name>"]["camera_fps"] }}'
        unit_of_measurement: 'FPS'
      <camera_name>_skipped_fps: 
        value_template: '{{ states.sensor.frigate_debug.attributes["<camera_name>"]["skipped_fps"] }}'
@@ -103,11 +544,11 @@ sensor:
      <camera_name>_detection_fps: 
        value_template: '{{ states.sensor.frigate_debug.attributes["<camera_name>"]["detection_fps"] }}'
        unit_of_measurement: 'FPS'
-      frigate_coral_fps: 
-        value_template: '{{ states.sensor.frigate_debug.attributes["coral"]["fps"] }}'
+      frigate_detection_fps: 
+        value_template: '{{ states.sensor.frigate_debug.attributes["detection_fps"] }}'
        unit_of_measurement: 'FPS'
      frigate_coral_inference:
-        value_template: '{{ states.sensor.frigate_debug.attributes["coral"]["inference_speed"] }}' 
+        value_template: '{{ states.sensor.frigate_debug.attributes["detectors"]["coral"]["inference_speed"] }}' 
        unit_of_measurement: 'ms'
        
 automation:
@@ -130,6 +571,9 @@ automation:
              - url: http://<ip>:5000/<camera_name>/person/best.jpg
                caption: A person was detected.        
 ```
+
+[Back to top](#documentation)
+
 ## HTTP Endpoints
 A web server is available on port 5000 with the following endpoints.

@@ -152,134 +596,9 @@ Example parameters:
 - `h=300`: resizes the image to 300 pixes tall

 ### `/debug/stats`
-Contains some granular debug info that can be used for sensors in HomeAssistant. See details below.
+Contains some granular debug info that can be used for sensors in HomeAssistant.

-## MQTT Messages
-These are the MQTT messages generated by Frigate. The default topic_prefix is `frigate`, but can be changed in the config file.
-
-### frigate/available
-Designed to be used as an availability topic with HomeAssistant. Possible message are:
-"online": published when frigate is running (on startup)
-"offline": published right before frigate stops
-
-### frigate/<camera_name>/<object_name>
-Publishes `ON` or `OFF` and is designed to be used a as a binary sensor in HomeAssistant for whether or not that object type is detected.
-
-### frigate/<camera_name>/<object_name>/snapshot
-Publishes a jpeg encoded frame of the detected object type. When the object is no longer detected, the highest confidence image is published or the original image
-is published again.
-
-The height and crop of snapshots can be configured as shown in the example config.
-
-### frigate/<camera_name>/events/start
-Message published at the start of any tracked object. JSON looks as follows:
-```json
-{
-    "label": "person",
-    "score": 0.87890625,
-    "box": [
-        95,
-        155,
-        581,
-        1182
-    ],
-    "area": 499122,
-    "region": [
-        0,
-        132,
-        1080,
-        1212
-    ],
-    "frame_time": 1600208805.60284,
-    "centroid": [
-        338,
-        668
-    ],
-    "id": "1600208805.60284-k1l43p",
-    "start_time": 1600208805.60284,
-    "top_score": 0.87890625,
-    "zones": [],
-    "score_history": [
-        0.87890625
-    ],
-    "computed_score": 0.0,
-    "false_positive": true
-}
-```
-
-### frigate/<camera_name>/events/end
-Same as `frigate/<camera_name>/events/start`, but with an `end_time` property as well.
-
-### frigate/<zone_name>/<object_name>
-Publishes `ON` or `OFF` and is designed to be used a as a binary sensor in HomeAssistant for whether or not that object type is detected in the zone.
-
-## Understanding min_score and threshold
-`min_score` defines the minimum score for Frigate to begin tracking a detected object. Any single detection below `min_score` will be ignored as a false positive. `threshold` is based on the median of the history of scores for a tracked object. Consider the following frames when `min_score` is set to 0.6 and threshold is set to 0.85:
-
-| Frame | Current Score | Score History | Computed Score | Detected Object |
-| --- | --- | --- | --- | --- |
-| 1 | 0.7 | 0.0, 0, 0.7 | 0.0 | No
-| 2 | 0.55 | 0.0, 0.7, 0.0 | 0.0 | No
-| 3 | 0.85 | 0.7, 0.0, 0.85 | 0.7 | No
-| 4 | 0.90 | 0.7, 0.85, 0.95, 0.90 | 0.875 | Yes
-| 5 | 0.88 | 0.7, 0.85, 0.95, 0.90, 0.88 | 0.88 | Yes
-| 6 | 0.95 | 0.7, 0.85, 0.95, 0.90, 0.88, 0.95 | 0.89 | Yes
-
-In frame 2, the score is below the `min_score` value, so frigate ignores it and it becomes a 0.0. The computed score is the median of the score history (padding to at least 3 values), and only when that computed score crosses the `threshold` is the object marked as a true positive. That happens in frame 4 in the example.
-
-## Using a custom model or labels
-Models for both CPU and EdgeTPU (Coral) are bundled in the image. You can use your own models with volume mounts:
- CPU Model: `/cpu_model.tflite`
- EdgeTPU Model: `/edgetpu_model.tflite`
- Labels: `/labelmap.txt`
-
-### Customizing the Labelmap
-The labelmap can be customized to your needs. A common reason to do this is to combine multiple object types that are easily confused when you don't need to be as granular such as car/truck. You must retain the same number of labels, but you can change the names. To change:
-
- Download the [COCO labelmap](https://dl.google.com/coral/canned_models/coco_labels.txt)
- Modify the label names as desired. For example, change `7 truck` to `7 car`
- Mount the new file at `/labelmap.txt` in the container with an additional volume
-  ```
-  -v ./config/labelmap.txt:/labelmap.txt
-  ```
-
-## Recording Clips
-**Note**: Previous versions of frigate included `-vsync drop` in input parameters. This is not compatible with FFmpeg's segment feature and must be removed from your input parameters if you have overrides set.
-
-Frigate can save video clips without any CPU overhead for encoding by simply copying the stream directly with FFmpeg. It leverages FFmpeg's segment functionality to maintain a cache of 90 seconds of video for each camera. The cache files are written to disk at /cache and do not introduce memory overhead. When an object is being tracked, it will extend the cache to ensure it can assemble a clip when the event ends. Once the event ends, it again uses FFmpeg to assemble a clip by combining the video clips without any encoding by the CPU. Assembled clips are are saved to the /clips directory along with a json file containing the current information about the tracked object.
-
-### Global Configuration Options
- `max_seconds`: This limits the size of the cache when an object is being tracked. If an object is stationary and being tracked for a long time, the cache files will expire and this value will be the maximum clip length for the *end* of the event. For example, if this is set to 300 seconds and an object is being tracked for 600 seconds, the clip will end up being the last 300 seconds. Defaults to 300 seconds.
-
-### Per-camera Configuration Options
- `pre_capture`: Defines how much time should be included in the clip prior to the beginning of the event. Defaults to 30 seconds.
- `objects`: List of object types to save clips for. Object types here must be listed for tracking at the camera or global configuration. Defaults to all tracked objects.
-
-## Google Coral Configuration
-Frigate attempts to detect your Coral device automatically. If you have multiple Coral devices or a version that is not detected automatically, you can specify using the `tensorflow_device` config option.
-
-## Masks and limiting detection to a certain area
-The mask works by looking at the bottom center of any bounding box (first image, red dot below) and comparing that to your mask. If that red dot falls on an area of your mask that is black, the detection (and motion) will be ignored. The mask in the second image would limit detection on this camera to only objects that are in the front yard and not the street. 
-
-<a href="docs/example-mask-check-point.png"><img src="docs/example-mask-check-point.png" height="300"></a>
-<a href="docs/example-mask.bmp"><img src="docs/example-mask.bmp" height="300"></a>
-<a href="docs/example-mask-overlay.png"><img src="docs/example-mask-overlay.png" height="300"></a>
-
-The following types of masks are supported:
- `base64`: Base64 encoded image file
- `poly`: List of x,y points like zone configuration
- `image`: Path to an image file in the config directory
-
-`base64` and `image` masks must be the same aspect ratio as your camera. 
-
-## Zones
-Zones allow you to define a specific area of the frame and apply additional filters for object types so you can determine whether or not an object is within a particular area. Zones cannot have the same name as a camera. If desired, a single zone can include multiple cameras if you have multiple cameras covering the same area. See the sample config for details on how to configure.
-
-During testing, `draw_zones` can be set in the config to tell frigate to draw the zone on the frames so you can adjust as needed. The zone line will increase in thickness when any object enters the zone.
-
-![Zone Example](docs/zone_example.jpg)
-
-## Debug Info
+Sample response:
 ```jsonc
 {
    /* Per Camera Stats */
@@ -336,43 +655,112 @@ During testing, `draw_zones` can be set in the config to tell frigate to draw th
        ***************/
        "skipped_fps": 0.0
    },
-    /* Coral Stats */
-    "coral": {
-        /***************
-        * Timestamp when object detection started. If this value stays non-zero and constant
-        * for a long time, that means the detection process is stuck.
-        ***************/
-        "detection_start": 0.0,
-        /***************
-        * Frames per second of the Coral. This should be the sum of all detection_fps values from cameras.
-        ***************/
-        "fps": 6.9,
-        /***************
-        * Time spent running object detection in milliseconds.
-        ***************/
-        "inference_speed": 10.48,
-        /***************
-        * PID for the shared process that runs object detection on the Coral.
-        ***************/
-        "pid": 25321
-    },
-    "plasma_store_rc": null // Return code for the plasma store. This should be null normally.
+    /***************
+    * Sum of detection_fps across all cameras and detectors. 
+    * This should be the sum of all detection_fps values from cameras.
+    ***************/
+    "detection_fps": 5.0,
+    /* Detectors Stats */
+    "detectors": {
+      "coral": {
+          /***************
+          * Timestamp when object detection started. If this value stays non-zero and constant
+          * for a long time, that means the detection process is stuck.
+          ***************/
+          "detection_start": 0.0,
+          /***************
+          * Time spent running object detection in milliseconds.
+          ***************/
+          "inference_speed": 10.48,
+          /***************
+          * PID for the shared process that runs object detection on the Coral.
+          ***************/
+          "pid": 25321
+      }
+    }
 }
 ```

-## Tips
- Lower the framerate of the video feed on the camera to reduce the CPU usage for capturing the feed. Not as effective, but you can also modify the `take_frame` [configuration](config/config.example.yml) for each camera to only analyze every other frame, or every third frame, etc. 
- Hard code the resolution of each camera in your config if you are having difficulty starting frigate or if the initial ffprobe for camerea resolution fails or returns incorrect info. Example:
+[Back to top](#documentation)
+
+## MQTT Topics
+These are the MQTT messages generated by Frigate. The default topic_prefix is `frigate`, but can be changed in the config file.
+
+### `frigate/available`
+Designed to be used as an availability topic with HomeAssistant. Possible message are:
+"online": published when frigate is running (on startup)
+"offline": published right before frigate stops
+
+### `frigate/<camera_name>/<object_name>`
+Publishes `ON` or `OFF` and is designed to be used a as a binary sensor in HomeAssistant for whether or not that object type is detected.
+
+### `frigate/<camera_name>/<object_name>/snapshot`
+Publishes a jpeg encoded frame of the detected object type. When the object is no longer detected, the highest confidence image is published or the original image
+is published again.
+
+The height and crop of snapshots can be configured in the config.
+
+### `frigate/<camera_name>/events/start`
+Message published at the start of any tracked object. JSON looks as follows:
+```json
+{
+    "label": "person",
+    "score": 0.87890625,
+    "box": [
+        95,
+        155,
+        581,
+        1182
+    ],
+    "area": 499122,
+    "region": [
+        0,
+        132,
+        1080,
+        1212
+    ],
+    "frame_time": 1600208805.60284,
+    "centroid": [
+        338,
+        668
+    ],
+    "id": "1600208805.60284-k1l43p",
+    "start_time": 1600208805.60284,
+    "top_score": 0.87890625,
+    "zones": [],
+    "score_history": [
+        0.87890625
+    ],
+    "computed_score": 0.0,
+    "false_positive": true
+}
 ```
-cameras:
-  back:
-    ffmpeg:
-      input: rtsp://<camera>
-    height: 1080
-    width: 1920
-```
- Additional logging is available in the docker container - You can view the logs by running `docker logs -t frigate`
- Object configuration - Tracked objects types, sizes and thresholds can be defined globally and/or on a per camera basis. The global and camera object configuration is *merged*. For example, if you defined tracking person, car, and truck globally but modified your backyard camera to only track person, the global config would merge making the effective list for the backyard camera still contain person, car and truck. If you want precise object tracking per camera, best practice to put a minimal list of objects at the global level and expand objects on a per camera basis. Object threshold and area configuration will be used first from the camera object config (if defined) and then from the global config.  See the [example config](config/config.example.yml) for more information.
+
+### `frigate/<camera_name>/events/end`
+Same as `frigate/<camera_name>/events/start`, but with an `end_time` property as well.
+
+### `frigate/<zone_name>/<object_name>`
+Publishes `ON` when the object enters the zone and `OFF` when the object disappears or exits the zone. Designed to be used a as a binary sensor in HomeAssistant for whether or not that object type is detected in the zone.
+
+[Back to top](#documentation)
+
+## Custom Models
+Models for both CPU and EdgeTPU (Coral) are bundled in the image. You can use your own models with volume mounts:
+- CPU Model: `/cpu_model.tflite`
+- EdgeTPU Model: `/edgetpu_model.tflite`
+- Labels: `/labelmap.txt`
+
+### Customizing the Labelmap
+The labelmap can be customized to your needs. A common reason to do this is to combine multiple object types that are easily confused when you don't need to be as granular such as car/truck. You must retain the same number of labels, but you can change the names. To change:
+
+- Download the [COCO labelmap](https://dl.google.com/coral/canned_models/coco_labels.txt)
+- Modify the label names as desired. For example, change `7 truck` to `7 car`
+- Mount the new file at `/labelmap.txt` in the container with an additional volume
+  ```
+  -v ./config/labelmap.txt:/labelmap.txt
+  ```
+
+[Back to top](#documentation)

 ## Troubleshooting

@@ -386,3 +774,8 @@ ffmpeg:
    - info
 ```

+### "On connect called"
+If you see repeated "On connect called" messages in your config, check for another instance of frigate. This happens when multiple frigate containers are trying to connect to mqtt with the same client_id.
+
+[Back to top](#documentation)
+
--- a/benchmark.py
+++ b/benchmark.py
@@ -11,7 +11,7 @@ labels = load_labels('/labelmap.txt')
 ######
 # Minimal same process runner
 ######
-# object_detector = ObjectDetector()
+# object_detector = LocalObjectDetector()
 # tensor_input = np.expand_dims(np.full((300,300,3), 0, np.uint8), axis=0)

 # start = datetime.datetime.now().timestamp()
@@ -37,11 +37,9 @@ labels = load_labels('/labelmap.txt')
 # print(f"Processed for {duration:.2f} seconds.")
 # print(f"Average frame processing time: {mean(frame_times)*1000:.2f}ms")

-######
-# Separate process runner
-######
-def start(id, num_detections, detection_queue):
-  object_detector = RemoteObjectDetector(str(id), '/labelmap.txt', detection_queue)
+
+def start(id, num_detections, detection_queue, event):
+  object_detector = RemoteObjectDetector(str(id), '/labelmap.txt', detection_queue, event)
  start = datetime.datetime.now().timestamp()

  frame_times = []
@@ -51,23 +49,39 @@ def start(id, num_detections, detection_queue):
    frame_times.append(datetime.datetime.now().timestamp()-start_frame)

  duration = datetime.datetime.now().timestamp()-start
+  object_detector.cleanup()
  print(f"{id} - Processed for {duration:.2f} seconds.")
+  print(f"{id} - FPS: {object_detector.fps.eps():.2f}")
  print(f"{id} - Average frame processing time: {mean(frame_times)*1000:.2f}ms")

-edgetpu_process = EdgeTPUProcess()
+######
+# Separate process runner
+######
+# event = mp.Event()
+# detection_queue = mp.Queue()
+# edgetpu_process = EdgeTPUProcess(detection_queue, {'1': event}, 'usb:0')

-# start(1, 1000, edgetpu_process.detect_lock, edgetpu_process.detect_ready, edgetpu_process.frame_ready)
+# start(1, 1000, edgetpu_process.detection_queue, event)
+# print(f"Average raw inference speed: {edgetpu_process.avg_inference_speed.value*1000:.2f}ms")

 ####
 # Multiple camera processes
 ####
 camera_processes = []
+
+events = {}
 for x in range(0, 10):
-  camera_process = mp.Process(target=start, args=(x, 100, edgetpu_process.detection_queue))
+  events[str(x)] = mp.Event()
+detection_queue = mp.Queue()
+edgetpu_process_1 = EdgeTPUProcess(detection_queue, events, 'usb:0')
+edgetpu_process_2 = EdgeTPUProcess(detection_queue, events, 'usb:1')
+
+for x in range(0, 10):
+  camera_process = mp.Process(target=start, args=(x, 300, detection_queue, events[str(x)]))
  camera_process.daemon = True
  camera_processes.append(camera_process)

-start = datetime.datetime.now().timestamp()
+start_time = datetime.datetime.now().timestamp()

 for p in camera_processes:
  p.start()
@@ -75,5 +89,5 @@ for p in camera_processes:
 for p in camera_processes:
  p.join()

-duration = datetime.datetime.now().timestamp()-start
+duration = datetime.datetime.now().timestamp()-start_time
 print(f"Total - Processed for {duration:.2f} seconds.")
--- a/config/back-mask.bmp
+++ b/config/back-mask.bmp
--- a/config/config.example.yml
+++ b/config/config.example.yml
@@ -1,227 +0,0 @@
-web_port: 5000
-
-################
-## Tell frigate to look for a specific EdgeTPU device. Useful if you want to run multiple instances of frigate
-## on the same machine with multiple EdgeTPUs. https://coral.ai/docs/edgetpu/multiple-edgetpu/#using-the-tensorflow-lite-python-api
-################
-tensorflow_device: usb
-
-mqtt:
-  host: mqtt.server.com
-  topic_prefix: frigate
-  # client_id: frigate # Optional -- set to override default client id of 'frigate' if running multiple instances
-  # user: username # Optional
-  #################
-  ## Environment variables that begin with 'FRIGATE_' may be referenced in {}.
-  ##   password: '{FRIGATE_MQTT_PASSWORD}'
-  #################
-  # password: password # Optional
-
-################
-# Global configuration for saving clips
-################
-save_clips:
-  ###########
-  # Maximum length of time to retain video during long events.
-  # If an object is being tracked for longer than this amount of time, the cache  
-  # will begin to expire and the resulting clip will be the last x seconds of the event.
-  ###########
-  max_seconds: 300
-
-#################
-# Default ffmpeg args. Optional and can be overwritten per camera.
-# Should work with most RTSP cameras that send h264 video
-# Built from the properties below with:
-# "ffmpeg" + global_args + input_args + "-i" + input + output_args
-#################
-# ffmpeg:
-#   global_args:
-#     - -hide_banner
-#     - -loglevel
-#     - panic
-#   hwaccel_args: []
-#   input_args:
-#     - -avoid_negative_ts
-#     - make_zero
-#     - -fflags
-#     - nobuffer
-#     - -flags
-#     - low_delay
-#     - -strict
-#     - experimental
-#     - -fflags
-#     - +genpts+discardcorrupt
-#     - -vsync
-#     - drop
-#     - -rtsp_transport
-#     - tcp
-#     - -stimeout
-#     - '5000000'
-#     - -use_wallclock_as_timestamps
-#     - '1'
-#   output_args:
-#     - -f
-#     - rawvideo
-#     - -pix_fmt
-#     - rgb24
-
-####################
-# Global object configuration. Applies to all cameras
-# unless overridden at the camera levels.
-# Keys must be valid labels. By default, the model uses coco (https://dl.google.com/coral/canned_models/coco_labels.txt).
-# All labels from the model are reported over MQTT. These values are used to filter out false positives.
-# min_area (optional): minimum width*height of the bounding box for the detected object
-# max_area (optional): maximum width*height of the bounding box for the detected object
-# min_score (optional): minimum score for the object to initiate tracking
-# threshold (optional): The minimum decimal percentage for tracked object's computed score to considered a true positive
-####################
-objects:
-  track:
-    - person
-  filters:
-    person:
-      min_area: 5000
-      max_area: 100000
-      min_score: 0.5
-      threshold: 0.85
-
-cameras:
-  back:
-    ffmpeg:
-      ################
-      # Source passed to ffmpeg after the -i parameter. Supports anything compatible with OpenCV and FFmpeg.
-      # Environment variables that begin with 'FRIGATE_' may be referenced in {}
-      ################
-      input: rtsp://viewer:{FRIGATE_RTSP_PASSWORD}@10.0.10.10:554/cam/realmonitor?channel=1&subtype=2
-      #################
-      # These values will override default values for just this camera
-      #################
-      # global_args: []
-      # hwaccel_args: []
-      # input_args: []
-      # output_args: []
-    
-    ################
-    ## Optionally specify the resolution of the video feed. Frigate will try to auto detect if not specified
-    ################
-    # height: 1280
-    # width: 720
-
-    ################
-    ## Specify the framerate of your camera
-    ##
-    ## NOTE: This should only be set in the event ffmpeg is unable to determine your camera's framerate
-    ##       on its own and the reported framerate for your camera in frigate is well over what is expected.
-    ################
-    # fps: 5
-
-    ################
-    ## Optional mask. Must be the same aspect ratio as your video feed. Value is any of the following:
-    ##   - name of a file in the config directory
-    ##   - base64 encoded image prefixed with 'base64,' eg. 'base64,asfasdfasdf....'
-    ##   - polygon of x,y coordinates prefixed with 'poly,' eg. 'poly,0,900,1080,900,1080,1920,0,1920'
-    ## 
-    ## The mask works by looking at the bottom center of the bounding box for the detected
-    ## person in the image. If that pixel in the mask is a black pixel, it ignores it as a
-    ## false positive. In my mask, the grass and driveway visible from my backdoor camera 
-    ## are white. The garage doors, sky, and trees (anywhere it would be impossible for a 
-    ## person to stand) are black.
-    ## 
-    ## Masked areas are also ignored for motion detection.
-    ################
-    # mask: back-mask.bmp
-
-    ################
-    # Allows you to limit the framerate within frigate for cameras that do not support
-    # custom framerates. A value of 1 tells frigate to look at every frame, 2 every 2nd frame, 
-    # 3 every 3rd frame, etc.
-    ################
-    take_frame: 1
-
-    ################
-    # The number of seconds to retain the highest scoring image for the best.jpg endpoint before allowing it
-    # to be replaced by a newer image. Defaults to 60 seconds.
-    ################
-    best_image_timeout: 60
-
-    ################
-    # MQTT settings
-    ################
-    # mqtt:
-    #   crop_to_region: True
-    #   snapshot_height: 300
-
-    ################
-    # Zones
-    ################
-    zones:
-      #################
-      # Name of the zone
-      ################
-      front_steps:
-        ####################
-        # A list of x,y coordinates to define the polygon of the zone. The top 
-        # left corner is 0,0. Can also be a comma separated string of all x,y coordinates combined.
-        # The same zone name can exist across multiple cameras if they have overlapping FOVs.
-        # An object is determined to be in the zone based on whether or not the bottom center
-        # of it's bounding box is within the polygon. The polygon must have at least 3 points.
-        # Coordinates can be generated at https://www.image-map.net/
-        ####################
-        coordinates:
-          - 545,1077
-          - 747,939
-          - 788,805
-        ################
-        # Zone level object filters. These are applied in addition to the global and camera filters
-        # and should be more restrictive than the global and camera filters. The global and camera
-        # filters are applied upstream.
-        ################
-        filters:
-          person:
-            min_area: 5000
-            max_area: 100000
-            threshold: 0.8
-
-    ################
-    # This will save a clip for each tracked object by frigate along with a json file that contains
-    # data related to the tracked object. This works by telling ffmpeg to write video segments to /cache
-    # from the video stream without re-encoding. Clips are then created by using ffmpeg to merge segments
-    # without re-encoding. The segments saved are unaltered from what frigate receives to avoid re-encoding.
-    # They do not contain bounding boxes. These are optimized to capture "false_positive" examples for improving frigate.
-    #
-    # NOTE: This feature does not work if you have "-vsync drop" configured in your input params. 
-    #       This will only work for camera feeds that can be copied into the mp4 container format without
-    #       encoding such as h264. It may not work for some types of streams.
-    ################
-    save_clips:
-      enabled: False
-      #########
-      # Number of seconds before the event to include in the clips
-      #########
-      pre_capture: 30
-      #########
-      # Objects to save clips for. Defaults to all tracked object types.
-      #########
-      # objects:
-      #   - person      
-
-    ################
-    # Configuration for the snapshots in the debug view and mqtt
-    ################
-    snapshots:
-      show_timestamp: True
-      draw_zones: False
-
-    ################
-    # Camera level object config. If defined, this is used instead of the global config.
-    ################
-    objects:
-      track:
-        - person
-        - car
-      filters:
-        person:
-          min_area: 5000
-          max_area: 100000
-          min_score: 0.5
-          threshold: 0.85
--- a/detect_objects.py
+++ b/detect_objects.py
@@ -1,3 +1,4 @@
+import faulthandler; faulthandler.enable()
 import os
 import signal
 import sys
@@ -8,6 +9,7 @@ import time
 import datetime
 import queue
 import yaml
+import json
 import threading
 import multiprocessing as mp
 import subprocess as sp
@@ -16,7 +18,7 @@ import logging
 from flask import Flask, Response, make_response, jsonify, request
 import paho.mqtt.client as mqtt

-from frigate.video import track_camera, get_ffmpeg_input, get_frame_shape, CameraCapture, start_or_restart_ffmpeg
+from frigate.video import capture_camera, track_camera, get_ffmpeg_input, get_frame_shape, CameraCapture, start_or_restart_ffmpeg
 from frigate.object_processing import TrackedObjectProcessor
 from frigate.events import EventProcessor
 from frigate.util import EventsPerSecond
@@ -24,8 +26,22 @@ from frigate.edgetpu import EdgeTPUProcess

 FRIGATE_VARS = {k: v for k, v in os.environ.items() if k.startswith('FRIGATE_')}

-with open('/config/config.yml') as f:
-    CONFIG = yaml.safe_load(f)
+CONFIG_FILE = os.environ.get('CONFIG_FILE', '/config/config.yml')
+
+if CONFIG_FILE.endswith(".yml"):
+    with open(CONFIG_FILE) as f:
+        CONFIG = yaml.safe_load(f)
+elif CONFIG_FILE.endswith(".json"):
+    with open(CONFIG_FILE) as f:
+        CONFIG = json.load(f)
+
+CACHE_DIR = CONFIG.get('save_clips', {}).get('cache_dir', '/cache')
+CLIPS_DIR = CONFIG.get('save_clips', {}).get('clips_dir', '/clips')
+
+if not os.path.exists(CACHE_DIR) and not os.path.islink(CACHE_DIR):
+    os.makedirs(CACHE_DIR)
+if not os.path.exists(CLIPS_DIR) and not os.path.islink(CLIPS_DIR):
+    os.makedirs(CLIPS_DIR)

 MQTT_HOST = CONFIG['mqtt']['host']
 MQTT_PORT = CONFIG.get('mqtt', {}).get('port', 1883)
@@ -54,32 +70,23 @@ FFMPEG_DEFAULT_CONFIG = {
         '-use_wallclock_as_timestamps', '1']),
    'output_args': FFMPEG_CONFIG.get('output_args',
        ['-f', 'rawvideo',
-         '-pix_fmt', 'rgb24'])
+         '-pix_fmt', 'yuv420p'])
 }

 GLOBAL_OBJECT_CONFIG = CONFIG.get('objects', {})

 WEB_PORT = CONFIG.get('web_port', 5000)
-DEBUG = (CONFIG.get('debug', '0') == '1')
-TENSORFLOW_DEVICE = CONFIG.get('tensorflow_device')
+DETECTORS = CONFIG.get('detectors', {'coral': {'type': 'edgetpu', 'device': 'usb'}})

-def start_plasma_store():
-    plasma_cmd = ['plasma_store', '-m', '400000000', '-s', '/tmp/plasma']
-    plasma_process = sp.Popen(plasma_cmd, stdout=sp.DEVNULL, stderr=sp.DEVNULL)
-    time.sleep(1)
-    rc = plasma_process.poll()
-    if rc is not None:
-        return None
-    return plasma_process
-
-class CameraWatchdog(threading.Thread):
-    def __init__(self, camera_processes, config, tflite_process, tracked_objects_queue, plasma_process, stop_event):
+class FrigateWatchdog(threading.Thread):
+    def __init__(self, camera_processes, config, detectors, detection_queue, out_events, tracked_objects_queue, stop_event):
        threading.Thread.__init__(self)
        self.camera_processes = camera_processes
        self.config = config
-        self.tflite_process = tflite_process
+        self.detectors = detectors
+        self.detection_queue = detection_queue
+        self.out_events = out_events
        self.tracked_objects_queue = tracked_objects_queue
-        self.plasma_process = plasma_process
        self.stop_event = stop_event

    def run(self):
@@ -93,60 +100,33 @@ class CameraWatchdog(threading.Thread):
                break

            now = datetime.datetime.now().timestamp()
-            
-            # check the plasma process
-            rc = self.plasma_process.poll()
-            if rc != None:
-                print(f"plasma_process exited unexpectedly with {rc}")
-                self.plasma_process = start_plasma_store()

-            # check the detection process
-            detection_start = self.tflite_process.detection_start.value
-            if (detection_start > 0.0 and 
-                now - detection_start > 10):
-                print("Detection appears to be stuck. Restarting detection process")
-                self.tflite_process.start_or_restart()
-            elif not self.tflite_process.detect_process.is_alive():
-                print("Detection appears to have stopped. Restarting detection process")
-                self.tflite_process.start_or_restart()
+            # check the detection processes
+            for detector in self.detectors.values():
+                detection_start = detector.detection_start.value
+                if (detection_start > 0.0 and 
+                    now - detection_start > 10):
+                    print("Detection appears to be stuck. Restarting detection process")
+                    detector.start_or_restart()
+                elif not detector.detect_process.is_alive():
+                    print("Detection appears to have stopped. Restarting detection process")
+                    detector.start_or_restart()

            # check the camera processes
            for name, camera_process in self.camera_processes.items():
                process = camera_process['process']
                if not process.is_alive():
                    print(f"Track process for {name} is not alive. Starting again...")
+                    camera_process['camera_fps'].value = 0.0
                    camera_process['process_fps'].value = 0.0
                    camera_process['detection_fps'].value = 0.0
                    camera_process['read_start'].value = 0.0
-                    process = mp.Process(target=track_camera, args=(name, self.config[name], camera_process['frame_queue'],
-                        camera_process['frame_shape'], self.tflite_process.detection_queue, self.tracked_objects_queue, 
-                        camera_process['process_fps'], camera_process['detection_fps'],
-                        camera_process['read_start'], camera_process['detection_frame'], self.stop_event))
+                    process = mp.Process(target=track_camera, args=(name, self.config,
+                        self.detection_queue, self.out_events[name], self.tracked_objects_queue, camera_process, self.stop_event))
                    process.daemon = True
                    camera_process['process'] = process
                    process.start()
                    print(f"Track process started for {name}: {process.pid}")
-                
-                if not camera_process['capture_thread'].is_alive():
-                    frame_shape = camera_process['frame_shape']
-                    frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
-                    ffmpeg_process = start_or_restart_ffmpeg(camera_process['ffmpeg_cmd'], frame_size)
-                    camera_capture = CameraCapture(name, ffmpeg_process, frame_shape, camera_process['frame_queue'], 
-                        camera_process['take_frame'], camera_process['camera_fps'], camera_process['detection_frame'], self.stop_event)
-                    camera_capture.start()
-                    camera_process['ffmpeg_process'] = ffmpeg_process
-                    camera_process['capture_thread'] = camera_capture
-                elif now - camera_process['capture_thread'].current_frame.value > 5:
-                    print(f"No frames received from {name} in 5 seconds. Exiting ffmpeg...")
-                    ffmpeg_process = camera_process['ffmpeg_process']
-                    ffmpeg_process.terminate()
-                    try:
-                        print("Waiting for ffmpeg to exit gracefully...")
-                        ffmpeg_process.communicate(timeout=30)
-                    except sp.TimeoutExpired:
-                        print("FFmpeg didnt exit. Force killing...")
-                        ffmpeg_process.kill()
-                        ffmpeg_process.communicate()

 def main():
    stop_event = threading.Event()
@@ -172,29 +152,44 @@ def main():
    client.connect(MQTT_HOST, MQTT_PORT, 60)
    client.loop_start()

-    plasma_process = start_plasma_store()
-
    ##
    # Setup config defaults for cameras
    ##
    for name, config in CONFIG['cameras'].items():
        config['snapshots'] = {
            'show_timestamp': config.get('snapshots', {}).get('show_timestamp', True),
-            'draw_zones': config.get('snapshots', {}).get('draw_zones', False)
+            'draw_zones': config.get('snapshots', {}).get('draw_zones', False),
+            'draw_bounding_boxes': config.get('snapshots', {}).get('draw_bounding_boxes', True)
        }
        config['zones'] = config.get('zones', {})

    # Queue for cameras to push tracked objects to
-    tracked_objects_queue = mp.Queue()
+    tracked_objects_queue = mp.Queue(maxsize=len(CONFIG['cameras'].keys())*2)

    # Queue for clip processing
    event_queue = mp.Queue()
-    
-    # Start the shared tflite process
-    tflite_process = EdgeTPUProcess(TENSORFLOW_DEVICE)

-    # start the camera processes
-    camera_processes = {}
+    # create the detection pipes and shms
+    out_events = {}
+    camera_shms = []
+    for name in CONFIG['cameras'].keys():
+        out_events[name] = mp.Event()
+        shm_in = mp.shared_memory.SharedMemory(name=name, create=True, size=300*300*3)
+        shm_out = mp.shared_memory.SharedMemory(name=f"out-{name}", create=True, size=20*6*4)
+        camera_shms.append(shm_in)
+        camera_shms.append(shm_out)
+
+    detection_queue = mp.Queue()
+
+    detectors = {}
+    for name, detector in DETECTORS.items():
+        if detector['type'] == 'cpu':
+            detectors[name] = EdgeTPUProcess(detection_queue, out_events=out_events, tf_device='cpu')
+        if detector['type'] == 'edgetpu':
+            detectors[name] = EdgeTPUProcess(detection_queue, out_events=out_events, tf_device=detector['device'])
+
+    # create the camera processes
+    camera_process_info = {}
    for name, config in CONFIG['cameras'].items():
        # Merge the ffmpeg config with the global config
        ffmpeg = config.get('ffmpeg', {})
@@ -222,7 +217,7 @@ def main():
                "-an",
                "-map",
                "0",
-                f"/cache/{name}-%Y%m%d%H%M%S.mp4"
+                f"{os.path.join(CACHE_DIR, name)}-%Y%m%d%H%M%S.mp4"
            ] + ffmpeg_output_args
        ffmpeg_cmd = (['ffmpeg'] +
                ffmpeg_global_args +
@@ -232,35 +227,24 @@ def main():
                ffmpeg_output_args +
                ['pipe:'])
        
+        config['ffmpeg_cmd'] = ffmpeg_cmd
+        
        if 'width' in config and 'height' in config:
            frame_shape = (config['height'], config['width'], 3)
        else:
            frame_shape = get_frame_shape(ffmpeg_input)
+        
+        config['frame_shape'] = frame_shape
+        config['take_frame'] = config.get('take_frame', 1)

-        frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
-        take_frame = config.get('take_frame', 1)
-
-        detection_frame = mp.Value('d', 0.0)
-
-        ffmpeg_process = start_or_restart_ffmpeg(ffmpeg_cmd, frame_size)
-        frame_queue = mp.Queue()
-        camera_fps = EventsPerSecond()
-        camera_fps.start()
-        camera_capture = CameraCapture(name, ffmpeg_process, frame_shape, frame_queue, take_frame, camera_fps, detection_frame, stop_event)
-        camera_capture.start()
-
-        camera_processes[name] = {
-            'camera_fps': camera_fps,
-            'take_frame': take_frame,
+        camera_process_info[name] = {
+            'camera_fps': mp.Value('d', 0.0),
+            'skipped_fps': mp.Value('d', 0.0),
            'process_fps': mp.Value('d', 0.0),
            'detection_fps': mp.Value('d', 0.0),
-            'detection_frame': detection_frame,
+            'detection_frame': mp.Value('d', 0.0),
            'read_start': mp.Value('d', 0.0),
-            'ffmpeg_process': ffmpeg_process,
-            'ffmpeg_cmd': ffmpeg_cmd,
-            'frame_queue': frame_queue,
-            'frame_shape': frame_shape,
-            'capture_thread': camera_capture
+            'frame_queue': mp.Queue(maxsize=2)
        }

        # merge global object config into camera object config
@@ -274,37 +258,44 @@ def main():
            'filters': object_filters
        }

-        camera_process = mp.Process(target=track_camera, args=(name, config, frame_queue, frame_shape,
-            tflite_process.detection_queue, tracked_objects_queue, camera_processes[name]['process_fps'], 
-            camera_processes[name]['detection_fps'], 
-            camera_processes[name]['read_start'], camera_processes[name]['detection_frame'], stop_event))
+        capture_process = mp.Process(target=capture_camera, args=(name, config,
+            camera_process_info[name], stop_event))
+        capture_process.daemon = True
+        camera_process_info[name]['capture_process'] = capture_process
+
+        camera_process = mp.Process(target=track_camera, args=(name, config,
+            detection_queue, out_events[name], tracked_objects_queue, camera_process_info[name], stop_event))
        camera_process.daemon = True
-        camera_processes[name]['process'] = camera_process
+        camera_process_info[name]['process'] = camera_process

-    for name, camera_process in camera_processes.items():
+    # start the camera_processes
+    for name, camera_process in camera_process_info.items():
+        camera_process['capture_process'].start()
+        print(f"Camera capture process started for {name}: {camera_process['capture_process'].pid}")
        camera_process['process'].start()
-        print(f"Camera_process started for {name}: {camera_process['process'].pid}")
+        print(f"Camera process started for {name}: {camera_process['process'].pid}")

-    event_processor = EventProcessor(CONFIG, camera_processes, '/cache', '/clips', event_queue, stop_event)
+    event_processor = EventProcessor(CONFIG, camera_process_info, CACHE_DIR, CLIPS_DIR, event_queue, stop_event)
    event_processor.start()
    
    object_processor = TrackedObjectProcessor(CONFIG['cameras'], client, MQTT_TOPIC_PREFIX, tracked_objects_queue, event_queue, stop_event)
    object_processor.start()
    
-    camera_watchdog = CameraWatchdog(camera_processes, CONFIG['cameras'], tflite_process, tracked_objects_queue, plasma_process, stop_event)
-    camera_watchdog.start()
+    frigate_watchdog = FrigateWatchdog(camera_process_info, CONFIG['cameras'], detectors, detection_queue, out_events, tracked_objects_queue, stop_event)
+    frigate_watchdog.start()

    def receiveSignal(signalNumber, frame):
        print('Received:', signalNumber)
        stop_event.set()
        event_processor.join()
        object_processor.join()
-        camera_watchdog.join()
-        for name, camera_process in camera_processes.items():
-            camera_process['capture_thread'].join()
-        rc = camera_watchdog.plasma_process.poll()
-        if rc == None:
-            camera_watchdog.plasma_process.terminate()
+        frigate_watchdog.join()
+
+        for detector in detectors.values():
+            detector.stop()
+        for shm in camera_shms:
+            shm.close()
+            shm.unlink()
        sys.exit()
    
    signal.signal(signal.SIGTERM, receiveSignal)
@@ -343,33 +334,29 @@ def main():

        total_detection_fps = 0

-        for name, camera_stats in camera_processes.items():
+        for name, camera_stats in camera_process_info.items():
            total_detection_fps += camera_stats['detection_fps'].value
-            capture_thread = camera_stats['capture_thread']
            stats[name] = {
-                'camera_fps': round(capture_thread.fps.eps(), 2),
+                'camera_fps': round(camera_stats['camera_fps'].value, 2),
                'process_fps': round(camera_stats['process_fps'].value, 2),
-                'skipped_fps': round(capture_thread.skipped_fps.eps(), 2),
+                'skipped_fps': round(camera_stats['skipped_fps'].value, 2),
                'detection_fps': round(camera_stats['detection_fps'].value, 2),
-                'read_start': camera_stats['read_start'].value,
                'pid': camera_stats['process'].pid,
-                'ffmpeg_pid': camera_stats['ffmpeg_process'].pid,
+                'capture_pid': camera_stats['capture_process'].pid,
                'frame_info': {
-                    'read': capture_thread.current_frame.value,
                    'detect': camera_stats['detection_frame'].value,
                    'process': object_processor.camera_data[name]['current_frame_time']
                }
            }
        
-        stats['coral'] = {
-            'fps': round(total_detection_fps, 2),
-            'inference_speed': round(tflite_process.avg_inference_speed.value*1000, 2),
-            'detection_start': tflite_process.detection_start.value,
-            'pid': tflite_process.detect_process.pid
-        }
-
-        rc = camera_watchdog.plasma_process.poll()
-        stats['plasma_store_rc'] = rc
+        stats['detectors'] = {}
+        for name, detector in detectors.items():
+            stats['detectors'][name] = {
+                'inference_speed': round(detector.avg_inference_speed.value*1000, 2),
+                'detection_start': detector.detection_start.value,
+                'pid': detector.detect_process.pid
+            }
+        stats['detection_fps'] = round(total_detection_fps, 2)

        return jsonify(stats)

@@ -377,7 +364,11 @@ def main():
    def best(camera_name, label):
        if camera_name in CONFIG['cameras']:
            best_object = object_processor.get_best(camera_name, label)
-            best_frame = best_object.get('frame', np.zeros((720,1280,3), np.uint8))
+            best_frame = best_object.get('frame')
+            if best_frame is None:
+                best_frame = np.zeros((720,1280,3), np.uint8)
+            else:
+                best_frame = cv2.cvtColor(best_frame, cv2.COLOR_YUV2BGR_I420)
            
            crop = bool(request.args.get('crop', 0, type=int))
            if crop:
@@ -388,7 +379,6 @@ def main():
            width = int(height*best_frame.shape[1]/best_frame.shape[0])

            best_frame = cv2.resize(best_frame, dsize=(width, height), interpolation=cv2.INTER_AREA)
-            best_frame = cv2.cvtColor(best_frame, cv2.COLOR_RGB2BGR)
            ret, jpg = cv2.imencode('.jpg', best_frame)
            response = make_response(jpg.tobytes())
            response.headers['Content-Type'] = 'image/jpg'
@@ -419,7 +409,6 @@ def main():
            width = int(height*frame.shape[1]/frame.shape[0])

            frame = cv2.resize(frame, dsize=(width, height), interpolation=cv2.INTER_AREA)
-            frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)

            ret, jpg = cv2.imencode('.jpg', frame)
            response = make_response(jpg.tobytes())
@@ -432,14 +421,12 @@ def main():
        while True:
            # max out at specified FPS
            time.sleep(1/fps)
-            frame = object_processor.get_current_frame(camera_name)
+            frame = object_processor.get_current_frame(camera_name, draw=True)
            if frame is None:
                frame = np.zeros((height,int(height*16/9),3), np.uint8)

            width = int(height*frame.shape[1]/frame.shape[0])
-
            frame = cv2.resize(frame, dsize=(width, height), interpolation=cv2.INTER_LINEAR)
-            frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)

            ret, jpg = cv2.imencode('.jpg', frame)
            yield (b'--frame\r\n'
@@ -448,8 +435,6 @@ def main():
    app.run(host='0.0.0.0', port=WEB_PORT, debug=False)

    object_processor.join()
-    
-    plasma_process.terminate()

 if __name__ == '__main__':
    main()
--- a/docker/Dockerfile.aarch64
+++ b/docker/Dockerfile.aarch64
@@ -0,0 +1,22 @@
+FROM frigate-base
+LABEL maintainer "blakeb@blakeshome.com"
+
+ENV DEBIAN_FRONTEND=noninteractive
+# Install packages for apt repo
+RUN apt-get -qq update \
+    && apt-get -qq install --no-install-recommends -y \
+        # ffmpeg runtime dependencies
+        libgomp1 \
+        # runtime dependencies
+        libopenexr24 \
+        libgstreamer1.0-0 \
+        libgstreamer-plugins-base1.0-0 \
+        libopenblas-base \
+        libjpeg-turbo8 \
+        libpng16-16 \
+        libtiff5 \
+        libdc1394-22 \
+    ## Tensorflow lite
+    && pip3 install https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp38-cp38-linux_aarch64.whl \
+    && rm -rf /var/lib/apt/lists/* \
+    && (apt-get autoremove -y; apt-get autoclean -y)
--- a/docker/Dockerfile.amd64
+++ b/docker/Dockerfile.amd64
@@ -0,0 +1,18 @@
+FROM frigate-base
+LABEL maintainer "blakeb@blakeshome.com"
+
+# By default, use the i965 driver
+ENV LIBVA_DRIVER_NAME=i965
+# Install packages for apt repo
+RUN apt-get -qq update \
+    && apt-get -qq install --no-install-recommends -y \
+        # ffmpeg dependencies
+        libgomp1 \
+        # VAAPI drivers for Intel hardware accel
+        libva-drm2 libva2 i965-va-driver vainfo intel-media-va-driver \
+        ## Tensorflow lite
+    && wget -q https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp38-cp38-linux_x86_64.whl \
+    && python3.8 -m pip install tflite_runtime-2.1.0.post1-cp38-cp38-linux_x86_64.whl \
+    && rm tflite_runtime-2.1.0.post1-cp38-cp38-linux_x86_64.whl \
+    && rm -rf /var/lib/apt/lists/* \
+    && (apt-get autoremove -y; apt-get autoclean -y)
--- a/docker/Dockerfile.armv7
+++ b/docker/Dockerfile.armv7
@@ -0,0 +1,24 @@
+FROM frigate-base
+LABEL maintainer "blakeb@blakeshome.com"
+
+ENV DEBIAN_FRONTEND=noninteractive
+# Install packages for apt repo
+RUN apt-get -qq update \
+    && apt-get -qq install --no-install-recommends -y \
+        # ffmpeg runtime dependencies
+        libgomp1 \
+        # runtime dependencies
+        libopenexr24 \
+        libgstreamer1.0-0 \
+        libgstreamer-plugins-base1.0-0 \
+        libopenblas-base \
+        libjpeg-turbo8 \
+        libpng16-16 \
+        libtiff5 \
+        libdc1394-22 \
+        libaom0 \
+        libx265-179 \
+    ## Tensorflow lite
+    && pip3 install https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp38-cp38-linux_armv7l.whl \
+    && rm -rf /var/lib/apt/lists/* \
+    && (apt-get autoremove -y; apt-get autoclean -y)
--- a/docker/Dockerfile.base
+++ b/docker/Dockerfile.base
@@ -0,0 +1,44 @@
+ARG ARCH=amd64
+FROM blakeblackshear/frigate-wheels:${ARCH} as wheels
+FROM blakeblackshear/frigate-ffmpeg:${ARCH} as ffmpeg
+
+FROM ubuntu:20.04
+LABEL maintainer "blakeb@blakeshome.com"
+
+COPY --from=ffmpeg /usr/local /usr/local/
+
+COPY --from=wheels /wheels/. /wheels/
+
+ENV FLASK_ENV=development
+ENV DEBIAN_FRONTEND=noninteractive
+# Install packages for apt repo
+RUN apt-get -qq update \
+    && apt-get upgrade -y \
+    && apt-get -qq install --no-install-recommends -y \
+    gnupg wget unzip tzdata \
+    && apt-get -qq install --no-install-recommends -y \
+        python3-pip \
+    && pip3 install -U /wheels/*.whl \
+    && APT_KEY_DONT_WARN_ON_DANGEROUS_USAGE=DontWarn apt-key adv --fetch-keys https://packages.cloud.google.com/apt/doc/apt-key.gpg \
+    && echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" > /etc/apt/sources.list.d/coral-edgetpu.list \
+    && echo "libedgetpu1-max libedgetpu/accepted-eula select true" | debconf-set-selections \
+    && apt-get -qq update && apt-get -qq install --no-install-recommends -y \
+        libedgetpu1-max \
+    && rm -rf /var/lib/apt/lists/* /wheels \
+    && (apt-get autoremove -y; apt-get autoclean -y)
+
+# get model and labels
+ARG MODEL_REFS=7064b94dd5b996189242320359dbab8b52c94a84
+COPY labelmap.txt /labelmap.txt
+RUN wget -q https://github.com/google-coral/edgetpu/raw/$MODEL_REFS/test_data/ssd_mobilenet_v2_coco_quant_postprocess_edgetpu.tflite -O /edgetpu_model.tflite
+RUN wget -q https://github.com/google-coral/edgetpu/raw/$MODEL_REFS/test_data/ssd_mobilenet_v2_coco_quant_postprocess.tflite -O /cpu_model.tflite
+
+RUN mkdir /cache /clips
+
+WORKDIR /opt/frigate/
+ADD frigate frigate/
+COPY detect_objects.py .
+COPY benchmark.py .
+COPY process_clip.py .
+
+CMD ["python3", "-u", "detect_objects.py"]
--- a/docker/Dockerfile.ffmpeg.aarch64
+++ b/docker/Dockerfile.ffmpeg.aarch64
@@ -0,0 +1,533 @@
+# inspired by:
+# https://github.com/collelog/ffmpeg/blob/master/4.3.1-alpine-rpi4-arm64v8.Dockerfile
+# https://github.com/mmastrac/ffmpeg-omx-rpi-docker/blob/master/Dockerfile
+# https://github.com/jrottenberg/ffmpeg/pull/158/files
+# https://github.com/jrottenberg/ffmpeg/pull/239
+FROM        ubuntu:20.04 AS base
+
+WORKDIR     /tmp/workdir
+
+ENV     DEBIAN_FRONTEND=noninteractive
+RUN     apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends ca-certificates expat libgomp1 && \
+        apt-get autoremove -y && \
+        apt-get clean -y
+
+FROM base as build
+
+ENV         FFMPEG_VERSION=4.3.1 \
+            AOM_VERSION=v1.0.0 \
+            FDKAAC_VERSION=0.1.5 \
+            FONTCONFIG_VERSION=2.12.4 \
+            FREETYPE_VERSION=2.5.5 \
+            FRIBIDI_VERSION=0.19.7 \
+            KVAZAAR_VERSION=1.2.0 \
+            LAME_VERSION=3.100 \
+            LIBASS_VERSION=0.13.7 \
+            LIBPTHREAD_STUBS_VERSION=0.4 \
+            LIBVIDSTAB_VERSION=1.1.0 \
+            LIBXCB_VERSION=1.13.1 \
+            XCBPROTO_VERSION=1.13 \
+            OGG_VERSION=1.3.2 \
+            OPENCOREAMR_VERSION=0.1.5 \
+            OPUS_VERSION=1.2 \
+            OPENJPEG_VERSION=2.1.2 \
+            THEORA_VERSION=1.1.1 \
+            VORBIS_VERSION=1.3.5 \
+            VPX_VERSION=1.8.0 \
+            WEBP_VERSION=1.0.2 \
+            X264_VERSION=20170226-2245-stable \
+            X265_VERSION=3.1.1 \
+            XAU_VERSION=1.0.9 \
+            XORG_MACROS_VERSION=1.19.2 \
+            XPROTO_VERSION=7.0.31 \
+            XVID_VERSION=1.3.4 \
+            LIBXML2_VERSION=2.9.10 \
+            LIBBLURAY_VERSION=1.1.2 \
+            LIBZMQ_VERSION=4.3.2 \
+            SRC=/usr/local
+
+ARG         FREETYPE_SHA256SUM="5d03dd76c2171a7601e9ce10551d52d4471cf92cd205948e60289251daddffa8 freetype-2.5.5.tar.gz"
+ARG         FRIBIDI_SHA256SUM="3fc96fa9473bd31dcb5500bdf1aa78b337ba13eb8c301e7c28923fea982453a8 0.19.7.tar.gz"
+ARG         LIBASS_SHA256SUM="8fadf294bf701300d4605e6f1d92929304187fca4b8d8a47889315526adbafd7 0.13.7.tar.gz"
+ARG         LIBVIDSTAB_SHA256SUM="14d2a053e56edad4f397be0cb3ef8eb1ec3150404ce99a426c4eb641861dc0bb v1.1.0.tar.gz"
+ARG         OGG_SHA256SUM="e19ee34711d7af328cb26287f4137e70630e7261b17cbe3cd41011d73a654692 libogg-1.3.2.tar.gz"
+ARG         OPUS_SHA256SUM="77db45a87b51578fbc49555ef1b10926179861d854eb2613207dc79d9ec0a9a9 opus-1.2.tar.gz"
+ARG         THEORA_SHA256SUM="40952956c47811928d1e7922cda3bc1f427eb75680c3c37249c91e949054916b libtheora-1.1.1.tar.gz"
+ARG         VORBIS_SHA256SUM="6efbcecdd3e5dfbf090341b485da9d176eb250d893e3eb378c428a2db38301ce libvorbis-1.3.5.tar.gz"
+ARG         XVID_SHA256SUM="4e9fd62728885855bc5007fe1be58df42e5e274497591fec37249e1052ae316f xvidcore-1.3.4.tar.gz"
+ARG         LIBXML2_SHA256SUM="f07dab13bf42d2b8db80620cce7419b3b87827cc937c8bb20fe13b8571ee9501  libxml2-v2.9.10.tar.gz"
+ARG         LIBBLURAY_SHA256SUM="a3dd452239b100dc9da0d01b30e1692693e2a332a7d29917bf84bb10ea7c0b42 libbluray-1.1.2.tar.bz2"
+ARG         LIBZMQ_SHA256SUM="02ecc88466ae38cf2c8d79f09cfd2675ba299a439680b64ade733e26a349edeb v4.3.2.tar.gz"
+
+
+ARG         LD_LIBRARY_PATH=/opt/ffmpeg/lib
+ARG         MAKEFLAGS="-j2"
+ARG         PKG_CONFIG_PATH="/opt/ffmpeg/share/pkgconfig:/opt/ffmpeg/lib/pkgconfig:/opt/ffmpeg/lib64/pkgconfig"
+ARG         PREFIX=/opt/ffmpeg
+ARG         LD_LIBRARY_PATH="/opt/ffmpeg/lib:/opt/ffmpeg/lib64:/usr/lib64:/usr/lib:/lib64:/lib"
+
+
+RUN      buildDeps="autoconf \
+                    automake \
+                    cmake \
+                    curl \
+                    bzip2 \
+                    libexpat1-dev \
+                    g++ \
+                    gcc \
+                    git \
+                    gperf \
+                    libtool \
+                    make \
+                    nasm \
+                    perl \
+                    pkg-config \
+                    python \
+                    libssl-dev \
+                    yasm \
+                    linux-headers-raspi2 \
+                    libomxil-bellagio-dev \
+                    zlib1g-dev" && \
+        apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends ${buildDeps}
+## opencore-amr https://sourceforge.net/projects/opencore-amr/
+RUN \
+        DIR=/tmp/opencore-amr && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://versaweb.dl.sourceforge.net/project/opencore-amr/opencore-amr/opencore-amr-${OPENCOREAMR_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## x264 http://www.videolan.org/developers/x264.html
+RUN \
+        DIR=/tmp/x264 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://download.videolan.org/pub/videolan/x264/snapshots/x264-snapshot-${X264_VERSION}.tar.bz2 | \
+        tar -jx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared --enable-pic --disable-cli && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### x265 http://x265.org/
+RUN \
+        DIR=/tmp/x265 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://download.videolan.org/pub/videolan/x265/x265_${X265_VERSION}.tar.gz  | \
+        tar -zx && \
+        cd x265_${X265_VERSION}/build/linux && \
+        sed -i "/-DEXTRA_LIB/ s/$/ -DCMAKE_INSTALL_PREFIX=\${PREFIX}/" multilib.sh && \
+        sed -i "/^cmake/ s/$/ -DENABLE_CLI=OFF/" multilib.sh && \
+        export CXXFLAGS="${CXXFLAGS} -fPIC" && \
+        ./multilib.sh && \
+        make -C 8bit install && \
+        rm -rf ${DIR}
+### libogg https://www.xiph.org/ogg/
+RUN \
+        DIR=/tmp/ogg && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/ogg/libogg-${OGG_VERSION}.tar.gz && \
+        echo ${OGG_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libogg-${OGG_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libopus https://www.opus-codec.org/
+RUN \
+        DIR=/tmp/opus && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://archive.mozilla.org/pub/opus/opus-${OPUS_VERSION}.tar.gz && \
+        echo ${OPUS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f opus-${OPUS_VERSION}.tar.gz && \
+        autoreconf -fiv && \
+        ./configure --prefix="${PREFIX}" --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libvorbis https://xiph.org/vorbis/
+RUN \
+        DIR=/tmp/vorbis && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/vorbis/libvorbis-${VORBIS_VERSION}.tar.gz && \
+        echo ${VORBIS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libvorbis-${VORBIS_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --with-ogg="${PREFIX}" --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libtheora http://www.theora.org/
+RUN \
+        DIR=/tmp/theora && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/theora/libtheora-${THEORA_VERSION}.tar.gz && \
+        echo ${THEORA_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libtheora-${THEORA_VERSION}.tar.gz && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD' -o config.guess && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD' -o config.sub && \
+        ./configure --prefix="${PREFIX}" --with-ogg="${PREFIX}" --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libvpx https://www.webmproject.org/code/
+RUN \
+        DIR=/tmp/vpx && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://codeload.github.com/webmproject/libvpx/tar.gz/v${VPX_VERSION} | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-vp8 --enable-vp9 --enable-vp9-highbitdepth --enable-pic --enable-shared \
+        --disable-debug --disable-examples --disable-docs --disable-install-bins  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libwebp https://developers.google.com/speed/webp/
+RUN \
+        DIR=/tmp/vebp && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-${WEBP_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libmp3lame http://lame.sourceforge.net/
+RUN \
+        DIR=/tmp/lame && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://versaweb.dl.sourceforge.net/project/lame/lame/$(echo ${LAME_VERSION} | sed -e 's/[^0-9]*\([0-9]*\)[.]\([0-9]*\)[.]\([0-9]*\)\([0-9A-Za-z-]*\)/\1.\2/')/lame-${LAME_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --bindir="${PREFIX}/bin" --enable-shared --enable-nasm --disable-frontend && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### xvid https://www.xvid.com/
+RUN \
+        DIR=/tmp/xvid && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xvid.org/downloads/xvidcore-${XVID_VERSION}.tar.gz && \
+        echo ${XVID_SHA256SUM} | sha256sum --check && \
+        tar -zx -f xvidcore-${XVID_VERSION}.tar.gz && \
+        cd xvidcore/build/generic && \
+        ./configure --prefix="${PREFIX}" --bindir="${PREFIX}/bin" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### fdk-aac https://github.com/mstorsjo/fdk-aac
+RUN \
+        DIR=/tmp/fdk-aac && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://github.com/mstorsjo/fdk-aac/archive/v${FDKAAC_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        autoreconf -fiv && \
+        ./configure --prefix="${PREFIX}" --enable-shared --datadir="${DIR}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## openjpeg https://github.com/uclouvain/openjpeg
+RUN \
+        DIR=/tmp/openjpeg && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://github.com/uclouvain/openjpeg/archive/v${OPENJPEG_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        export CFLAGS="${CFLAGS} -DPNG_ARM_NEON_OPT=0" && \
+        cmake -DBUILD_THIRDPARTY:BOOL=ON -DCMAKE_INSTALL_PREFIX="${PREFIX}" . && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## freetype https://www.freetype.org/
+RUN  \
+        DIR=/tmp/freetype && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://download.savannah.gnu.org/releases/freetype/freetype-${FREETYPE_VERSION}.tar.gz && \
+        echo ${FREETYPE_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f freetype-${FREETYPE_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## libvstab https://github.com/georgmartius/vid.stab
+RUN  \
+        DIR=/tmp/vid.stab && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/georgmartius/vid.stab/archive/v${LIBVIDSTAB_VERSION}.tar.gz && \
+        echo ${LIBVIDSTAB_SHA256SUM} | sha256sum --check &&  \
+        tar -zx --strip-components=1 -f v${LIBVIDSTAB_VERSION}.tar.gz && \
+        cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" . && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## fridibi https://www.fribidi.org/
+RUN  \
+        DIR=/tmp/fribidi && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/fribidi/fribidi/archive/${FRIBIDI_VERSION}.tar.gz && \
+        echo ${FRIBIDI_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f ${FRIBIDI_VERSION}.tar.gz && \
+        sed -i 's/^SUBDIRS =.*/SUBDIRS=gen.tab charset lib bin/' Makefile.am && \
+        ./bootstrap --no-config --auto && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j1 && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## fontconfig https://www.freedesktop.org/wiki/Software/fontconfig/
+RUN  \
+        DIR=/tmp/fontconfig && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.freedesktop.org/software/fontconfig/release/fontconfig-${FONTCONFIG_VERSION}.tar.bz2 && \
+        tar -jx --strip-components=1 -f fontconfig-${FONTCONFIG_VERSION}.tar.bz2 && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## libass https://github.com/libass/libass
+RUN  \
+        DIR=/tmp/libass && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/libass/libass/archive/${LIBASS_VERSION}.tar.gz && \
+        echo ${LIBASS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f ${LIBASS_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## kvazaar https://github.com/ultravideo/kvazaar
+RUN \
+        DIR=/tmp/kvazaar && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/ultravideo/kvazaar/archive/v${KVAZAAR_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f v${KVAZAAR_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/aom && \
+        git clone --branch ${AOM_VERSION} --depth 1 https://aomedia.googlesource.com/aom ${DIR} ; \
+        cd ${DIR} ; \
+        rm -rf CMakeCache.txt CMakeFiles ; \
+        mkdir -p ./aom_build ; \
+        cd ./aom_build ; \
+        cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" -DBUILD_SHARED_LIBS=1 ..; \
+        make ; \
+        make install ; \
+        rm -rf ${DIR}
+
+## libxcb (and supporting libraries) for screen capture https://xcb.freedesktop.org/
+RUN \
+        DIR=/tmp/xorg-macros && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive//individual/util/util-macros-${XORG_MACROS_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f util-macros-${XORG_MACROS_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/xproto && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive/individual/proto/xproto-${XPROTO_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f xproto-${XPROTO_VERSION}.tar.gz && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD' -o config.guess && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD' -o config.sub && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libXau && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive/individual/lib/libXau-${XAU_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libXau-${XAU_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libpthread-stubs && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/libpthread-stubs-${LIBPTHREAD_STUBS_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libpthread-stubs-${LIBPTHREAD_STUBS_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libxcb-proto && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/xcb-proto-${XCBPROTO_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f xcb-proto-${XCBPROTO_VERSION}.tar.gz && \
+        ACLOCAL_PATH="${PREFIX}/share/aclocal" ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libxcb && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/libxcb-${LIBXCB_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libxcb-${LIBXCB_VERSION}.tar.gz && \
+        ACLOCAL_PATH="${PREFIX}/share/aclocal" ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## libxml2 - for libbluray
+RUN \
+        DIR=/tmp/libxml2 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://gitlab.gnome.org/GNOME/libxml2/-/archive/v${LIBXML2_VERSION}/libxml2-v${LIBXML2_VERSION}.tar.gz && \
+        echo ${LIBXML2_SHA256SUM} | sha256sum --check && \
+        tar -xz --strip-components=1 -f libxml2-v${LIBXML2_VERSION}.tar.gz && \
+        ./autogen.sh --prefix="${PREFIX}" --with-ftp=no --with-http=no --with-python=no && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## libbluray - Requires libxml, freetype, and fontconfig
+RUN \
+        DIR=/tmp/libbluray && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://download.videolan.org/pub/videolan/libbluray/${LIBBLURAY_VERSION}/libbluray-${LIBBLURAY_VERSION}.tar.bz2 && \
+        echo ${LIBBLURAY_SHA256SUM} | sha256sum --check && \
+        tar -jx --strip-components=1 -f libbluray-${LIBBLURAY_VERSION}.tar.bz2 && \
+        ./configure --prefix="${PREFIX}" --disable-examples --disable-bdjava-jar --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## libzmq https://github.com/zeromq/libzmq/
+RUN \
+        DIR=/tmp/libzmq && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/zeromq/libzmq/archive/v${LIBZMQ_VERSION}.tar.gz && \
+        echo ${LIBZMQ_SHA256SUM} | sha256sum --check && \
+        tar -xz --strip-components=1 -f v${LIBZMQ_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make check && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## ffmpeg https://ffmpeg.org/
+RUN  \
+        DIR=/tmp/ffmpeg && mkdir -p ${DIR} && cd ${DIR} && \
+        curl -sLO https://ffmpeg.org/releases/ffmpeg-${FFMPEG_VERSION}.tar.bz2 && \
+        tar -jx --strip-components=1 -f ffmpeg-${FFMPEG_VERSION}.tar.bz2
+
+RUN \
+        DIR=/tmp/ffmpeg && mkdir -p ${DIR} && cd ${DIR} && \
+        ./configure \
+        --disable-debug \
+        --disable-doc \
+        --disable-ffplay \
+        --enable-shared \
+        --enable-avresample \
+        --enable-libopencore-amrnb \
+        --enable-libopencore-amrwb \
+        --enable-gpl \
+        --enable-libass \
+        --enable-fontconfig \
+        --enable-libfreetype \
+        --enable-libvidstab \
+        --enable-libmp3lame \
+        --enable-libopus \
+        --enable-libtheora \
+        --enable-libvorbis \
+        --enable-libvpx \
+        --enable-libwebp \
+        --enable-libxcb \
+        --enable-libx265 \
+        --enable-libxvid \
+        --enable-libx264 \
+        --enable-nonfree \
+        --enable-openssl \
+        --enable-libfdk_aac \
+        --enable-postproc \
+        --enable-small \
+        --enable-version3 \
+        --enable-libbluray \
+        --enable-libzmq \
+        --extra-libs=-ldl \
+        --prefix="${PREFIX}" \
+        --enable-libopenjpeg \
+        --enable-libkvazaar \
+        --enable-libaom \
+        --extra-libs=-lpthread \
+        # --enable-omx \
+        # --enable-omx-rpi \
+        # --enable-mmal \
+        --enable-v4l2_m2m \
+        --enable-neon \
+        --extra-cflags="-I${PREFIX}/include" \
+        --extra-ldflags="-L${PREFIX}/lib" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        make tools/zmqsend && cp tools/zmqsend ${PREFIX}/bin/ && \
+        make distclean && \
+        hash -r && \
+        cd tools && \
+        make qt-faststart && cp qt-faststart ${PREFIX}/bin/
+
+## cleanup
+RUN \
+        ldd ${PREFIX}/bin/ffmpeg | grep opt/ffmpeg | cut -d ' ' -f 3 | xargs -i cp {} /usr/local/lib/ && \
+        for lib in /usr/local/lib/*.so.*; do ln -s "${lib##*/}" "${lib%%.so.*}".so; done && \
+        cp ${PREFIX}/bin/* /usr/local/bin/ && \
+        cp -r ${PREFIX}/share/ffmpeg /usr/local/share/ && \
+        LD_LIBRARY_PATH=/usr/local/lib ffmpeg -buildconf && \
+        cp -r ${PREFIX}/include/libav* ${PREFIX}/include/libpostproc ${PREFIX}/include/libsw* /usr/local/include && \
+        mkdir -p /usr/local/lib/pkgconfig && \
+        for pc in ${PREFIX}/lib/pkgconfig/libav*.pc ${PREFIX}/lib/pkgconfig/libpostproc.pc ${PREFIX}/lib/pkgconfig/libsw*.pc; do \
+          sed "s:${PREFIX}:/usr/local:g" <"$pc" >/usr/local/lib/pkgconfig/"${pc##*/}"; \
+        done
+
+FROM        base AS release
+
+ENV         LD_LIBRARY_PATH=/usr/local/lib:/usr/local/lib64:/usr/lib:/usr/lib64:/lib:/lib64
+
+CMD         ["--help"]
+ENTRYPOINT  ["ffmpeg"]
+
+COPY --from=build /usr/local /usr/local/
+
+# Run ffmpeg with -c:v h264_v4l2m2m to enable HW accell for decoding on raspberry pi4 64-bit
--- a/docker/Dockerfile.ffmpeg.amd64
+++ b/docker/Dockerfile.ffmpeg.amd64
@@ -0,0 +1,526 @@
+# inspired by:
+# https://github.com/collelog/ffmpeg/blob/master/4.3.1-alpine-rpi4-arm64v8.Dockerfile
+# https://github.com/jrottenberg/ffmpeg/pull/158/files
+# https://github.com/jrottenberg/ffmpeg/pull/239
+FROM        ubuntu:20.04 AS base
+
+WORKDIR     /tmp/workdir
+
+ENV     DEBIAN_FRONTEND=noninteractive
+RUN     apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends ca-certificates expat libgomp1 && \
+        apt-get autoremove -y && \
+        apt-get clean -y
+
+FROM base as build
+
+ENV         FFMPEG_VERSION=4.3.1 \
+            AOM_VERSION=v1.0.0 \
+            FDKAAC_VERSION=0.1.5 \
+            FONTCONFIG_VERSION=2.12.4 \
+            FREETYPE_VERSION=2.5.5 \
+            FRIBIDI_VERSION=0.19.7 \
+            KVAZAAR_VERSION=1.2.0 \
+            LAME_VERSION=3.100 \
+            LIBASS_VERSION=0.13.7 \
+            LIBPTHREAD_STUBS_VERSION=0.4 \
+            LIBVIDSTAB_VERSION=1.1.0 \
+            LIBXCB_VERSION=1.13.1 \
+            XCBPROTO_VERSION=1.13 \
+            OGG_VERSION=1.3.2 \
+            OPENCOREAMR_VERSION=0.1.5 \
+            OPUS_VERSION=1.2 \
+            OPENJPEG_VERSION=2.1.2 \
+            THEORA_VERSION=1.1.1 \
+            VORBIS_VERSION=1.3.5 \
+            VPX_VERSION=1.8.0 \
+            WEBP_VERSION=1.0.2 \
+            X264_VERSION=20170226-2245-stable \
+            X265_VERSION=3.1.1 \
+            XAU_VERSION=1.0.9 \
+            XORG_MACROS_VERSION=1.19.2 \
+            XPROTO_VERSION=7.0.31 \
+            XVID_VERSION=1.3.4 \
+            LIBXML2_VERSION=2.9.10 \
+            LIBBLURAY_VERSION=1.1.2 \
+            LIBZMQ_VERSION=4.3.2 \
+            SRC=/usr/local
+
+ARG         FREETYPE_SHA256SUM="5d03dd76c2171a7601e9ce10551d52d4471cf92cd205948e60289251daddffa8 freetype-2.5.5.tar.gz"
+ARG         FRIBIDI_SHA256SUM="3fc96fa9473bd31dcb5500bdf1aa78b337ba13eb8c301e7c28923fea982453a8 0.19.7.tar.gz"
+ARG         LIBASS_SHA256SUM="8fadf294bf701300d4605e6f1d92929304187fca4b8d8a47889315526adbafd7 0.13.7.tar.gz"
+ARG         LIBVIDSTAB_SHA256SUM="14d2a053e56edad4f397be0cb3ef8eb1ec3150404ce99a426c4eb641861dc0bb v1.1.0.tar.gz"
+ARG         OGG_SHA256SUM="e19ee34711d7af328cb26287f4137e70630e7261b17cbe3cd41011d73a654692 libogg-1.3.2.tar.gz"
+ARG         OPUS_SHA256SUM="77db45a87b51578fbc49555ef1b10926179861d854eb2613207dc79d9ec0a9a9 opus-1.2.tar.gz"
+ARG         THEORA_SHA256SUM="40952956c47811928d1e7922cda3bc1f427eb75680c3c37249c91e949054916b libtheora-1.1.1.tar.gz"
+ARG         VORBIS_SHA256SUM="6efbcecdd3e5dfbf090341b485da9d176eb250d893e3eb378c428a2db38301ce libvorbis-1.3.5.tar.gz"
+ARG         XVID_SHA256SUM="4e9fd62728885855bc5007fe1be58df42e5e274497591fec37249e1052ae316f xvidcore-1.3.4.tar.gz"
+ARG         LIBXML2_SHA256SUM="f07dab13bf42d2b8db80620cce7419b3b87827cc937c8bb20fe13b8571ee9501  libxml2-v2.9.10.tar.gz"
+ARG         LIBBLURAY_SHA256SUM="a3dd452239b100dc9da0d01b30e1692693e2a332a7d29917bf84bb10ea7c0b42 libbluray-1.1.2.tar.bz2"
+ARG         LIBZMQ_SHA256SUM="02ecc88466ae38cf2c8d79f09cfd2675ba299a439680b64ade733e26a349edeb v4.3.2.tar.gz"
+
+
+ARG         LD_LIBRARY_PATH=/opt/ffmpeg/lib
+ARG         MAKEFLAGS="-j2"
+ARG         PKG_CONFIG_PATH="/opt/ffmpeg/share/pkgconfig:/opt/ffmpeg/lib/pkgconfig:/opt/ffmpeg/lib64/pkgconfig"
+ARG         PREFIX=/opt/ffmpeg
+ARG         LD_LIBRARY_PATH="/opt/ffmpeg/lib:/opt/ffmpeg/lib64:/usr/lib64:/usr/lib:/lib64:/lib"
+
+
+RUN      buildDeps="autoconf \
+                    automake \
+                    cmake \
+                    curl \
+                    bzip2 \
+                    libexpat1-dev \
+                    g++ \
+                    gcc \
+                    git \
+                    gperf \
+                    libtool \
+                    make \
+                    nasm \
+                    perl \
+                    pkg-config \
+                    python \
+                    libssl-dev \
+                    yasm \
+                    libva-dev \
+                    zlib1g-dev" && \
+        apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends ${buildDeps}
+## opencore-amr https://sourceforge.net/projects/opencore-amr/
+RUN \
+        DIR=/tmp/opencore-amr && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://versaweb.dl.sourceforge.net/project/opencore-amr/opencore-amr/opencore-amr-${OPENCOREAMR_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## x264 http://www.videolan.org/developers/x264.html
+RUN \
+        DIR=/tmp/x264 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://download.videolan.org/pub/videolan/x264/snapshots/x264-snapshot-${X264_VERSION}.tar.bz2 | \
+        tar -jx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared --enable-pic --disable-cli && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### x265 http://x265.org/
+RUN \
+        DIR=/tmp/x265 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://download.videolan.org/pub/videolan/x265/x265_${X265_VERSION}.tar.gz  | \
+        tar -zx && \
+        cd x265_${X265_VERSION}/build/linux && \
+        sed -i "/-DEXTRA_LIB/ s/$/ -DCMAKE_INSTALL_PREFIX=\${PREFIX}/" multilib.sh && \
+        sed -i "/^cmake/ s/$/ -DENABLE_CLI=OFF/" multilib.sh && \
+        ./multilib.sh && \
+        make -C 8bit install && \
+        rm -rf ${DIR}
+### libogg https://www.xiph.org/ogg/
+RUN \
+        DIR=/tmp/ogg && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/ogg/libogg-${OGG_VERSION}.tar.gz && \
+        echo ${OGG_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libogg-${OGG_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### libopus https://www.opus-codec.org/
+RUN \
+        DIR=/tmp/opus && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://archive.mozilla.org/pub/opus/opus-${OPUS_VERSION}.tar.gz && \
+        echo ${OPUS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f opus-${OPUS_VERSION}.tar.gz && \
+        autoreconf -fiv && \
+        ./configure --prefix="${PREFIX}" --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### libvorbis https://xiph.org/vorbis/
+RUN \
+        DIR=/tmp/vorbis && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/vorbis/libvorbis-${VORBIS_VERSION}.tar.gz && \
+        echo ${VORBIS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libvorbis-${VORBIS_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --with-ogg="${PREFIX}" --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### libtheora http://www.theora.org/
+RUN \
+        DIR=/tmp/theora && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/theora/libtheora-${THEORA_VERSION}.tar.gz && \
+        echo ${THEORA_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libtheora-${THEORA_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --with-ogg="${PREFIX}" --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### libvpx https://www.webmproject.org/code/
+RUN \
+        DIR=/tmp/vpx && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://codeload.github.com/webmproject/libvpx/tar.gz/v${VPX_VERSION} | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-vp8 --enable-vp9 --enable-vp9-highbitdepth --enable-pic --enable-shared \
+        --disable-debug --disable-examples --disable-docs --disable-install-bins  && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### libwebp https://developers.google.com/speed/webp/
+RUN \
+        DIR=/tmp/vebp && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-${WEBP_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### libmp3lame http://lame.sourceforge.net/
+RUN \
+        DIR=/tmp/lame && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://versaweb.dl.sourceforge.net/project/lame/lame/$(echo ${LAME_VERSION} | sed -e 's/[^0-9]*\([0-9]*\)[.]\([0-9]*\)[.]\([0-9]*\)\([0-9A-Za-z-]*\)/\1.\2/')/lame-${LAME_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --bindir="${PREFIX}/bin" --enable-shared --enable-nasm --disable-frontend && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### xvid https://www.xvid.com/
+RUN \
+        DIR=/tmp/xvid && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xvid.org/downloads/xvidcore-${XVID_VERSION}.tar.gz && \
+        echo ${XVID_SHA256SUM} | sha256sum --check && \
+        tar -zx -f xvidcore-${XVID_VERSION}.tar.gz && \
+        cd xvidcore/build/generic && \
+        ./configure --prefix="${PREFIX}" --bindir="${PREFIX}/bin" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+### fdk-aac https://github.com/mstorsjo/fdk-aac
+RUN \
+        DIR=/tmp/fdk-aac && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://github.com/mstorsjo/fdk-aac/archive/v${FDKAAC_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        autoreconf -fiv && \
+        ./configure --prefix="${PREFIX}" --enable-shared --datadir="${DIR}" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## openjpeg https://github.com/uclouvain/openjpeg
+RUN \
+        DIR=/tmp/openjpeg && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://github.com/uclouvain/openjpeg/archive/v${OPENJPEG_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        cmake -DBUILD_THIRDPARTY:BOOL=ON -DCMAKE_INSTALL_PREFIX="${PREFIX}" . && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## freetype https://www.freetype.org/
+RUN  \
+        DIR=/tmp/freetype && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://download.savannah.gnu.org/releases/freetype/freetype-${FREETYPE_VERSION}.tar.gz && \
+        echo ${FREETYPE_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f freetype-${FREETYPE_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## libvstab https://github.com/georgmartius/vid.stab
+RUN  \
+        DIR=/tmp/vid.stab && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/georgmartius/vid.stab/archive/v${LIBVIDSTAB_VERSION}.tar.gz && \
+        echo ${LIBVIDSTAB_SHA256SUM} | sha256sum --check &&  \
+        tar -zx --strip-components=1 -f v${LIBVIDSTAB_VERSION}.tar.gz && \
+        cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" . && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## fridibi https://www.fribidi.org/
+RUN  \
+        DIR=/tmp/fribidi && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/fribidi/fribidi/archive/${FRIBIDI_VERSION}.tar.gz && \
+        echo ${FRIBIDI_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f ${FRIBIDI_VERSION}.tar.gz && \
+        sed -i 's/^SUBDIRS =.*/SUBDIRS=gen.tab charset lib bin/' Makefile.am && \
+        ./bootstrap --no-config --auto && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j1 && \
+        make install && \
+        rm -rf ${DIR}
+## fontconfig https://www.freedesktop.org/wiki/Software/fontconfig/
+RUN  \
+        DIR=/tmp/fontconfig && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.freedesktop.org/software/fontconfig/release/fontconfig-${FONTCONFIG_VERSION}.tar.bz2 && \
+        tar -jx --strip-components=1 -f fontconfig-${FONTCONFIG_VERSION}.tar.bz2 && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## libass https://github.com/libass/libass
+RUN  \
+        DIR=/tmp/libass && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/libass/libass/archive/${LIBASS_VERSION}.tar.gz && \
+        echo ${LIBASS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f ${LIBASS_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+## kvazaar https://github.com/ultravideo/kvazaar
+RUN \
+        DIR=/tmp/kvazaar && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/ultravideo/kvazaar/archive/v${KVAZAAR_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f v${KVAZAAR_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/aom && \
+        git clone --branch ${AOM_VERSION} --depth 1 https://aomedia.googlesource.com/aom ${DIR} ; \
+        cd ${DIR} ; \
+        rm -rf CMakeCache.txt CMakeFiles ; \
+        mkdir -p ./aom_build ; \
+        cd ./aom_build ; \
+        cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" -DBUILD_SHARED_LIBS=1 ..; \
+        make ; \
+        make install ; \
+        rm -rf ${DIR}
+
+## libxcb (and supporting libraries) for screen capture https://xcb.freedesktop.org/
+RUN \
+        DIR=/tmp/xorg-macros && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive//individual/util/util-macros-${XORG_MACROS_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f util-macros-${XORG_MACROS_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/xproto && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive/individual/proto/xproto-${XPROTO_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f xproto-${XPROTO_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libXau && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive/individual/lib/libXau-${XAU_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libXau-${XAU_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libpthread-stubs && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/libpthread-stubs-${LIBPTHREAD_STUBS_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libpthread-stubs-${LIBPTHREAD_STUBS_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libxcb-proto && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/xcb-proto-${XCBPROTO_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f xcb-proto-${XCBPROTO_VERSION}.tar.gz && \
+        ACLOCAL_PATH="${PREFIX}/share/aclocal" ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libxcb && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/libxcb-${LIBXCB_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libxcb-${LIBXCB_VERSION}.tar.gz && \
+        ACLOCAL_PATH="${PREFIX}/share/aclocal" ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+## libxml2 - for libbluray
+RUN \
+        DIR=/tmp/libxml2 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://gitlab.gnome.org/GNOME/libxml2/-/archive/v${LIBXML2_VERSION}/libxml2-v${LIBXML2_VERSION}.tar.gz && \
+        echo ${LIBXML2_SHA256SUM} | sha256sum --check && \
+        tar -xz --strip-components=1 -f libxml2-v${LIBXML2_VERSION}.tar.gz && \
+        ./autogen.sh --prefix="${PREFIX}" --with-ftp=no --with-http=no --with-python=no && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+## libbluray - Requires libxml, freetype, and fontconfig
+RUN \
+        DIR=/tmp/libbluray && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://download.videolan.org/pub/videolan/libbluray/${LIBBLURAY_VERSION}/libbluray-${LIBBLURAY_VERSION}.tar.bz2 && \
+        echo ${LIBBLURAY_SHA256SUM} | sha256sum --check && \
+        tar -jx --strip-components=1 -f libbluray-${LIBBLURAY_VERSION}.tar.bz2 && \
+        ./configure --prefix="${PREFIX}" --disable-examples --disable-bdjava-jar --disable-static --enable-shared && \
+        make && \
+        make install && \
+        rm -rf ${DIR}
+
+## libzmq https://github.com/zeromq/libzmq/
+RUN \
+        DIR=/tmp/libzmq && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/zeromq/libzmq/archive/v${LIBZMQ_VERSION}.tar.gz && \
+        echo ${LIBZMQ_SHA256SUM} | sha256sum --check && \
+        tar -xz --strip-components=1 -f v${LIBZMQ_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" && \
+        make && \
+        make check && \
+        make install && \
+        rm -rf ${DIR}
+
+## ffmpeg https://ffmpeg.org/
+RUN  \
+        DIR=/tmp/ffmpeg && mkdir -p ${DIR} && cd ${DIR} && \
+        curl -sLO https://ffmpeg.org/releases/ffmpeg-${FFMPEG_VERSION}.tar.bz2 && \
+        tar -jx --strip-components=1 -f ffmpeg-${FFMPEG_VERSION}.tar.bz2
+
+
+
+RUN \
+        DIR=/tmp/ffmpeg && mkdir -p ${DIR} && cd ${DIR} && \
+        ./configure \
+        --disable-debug \
+        --disable-doc \
+        --disable-ffplay \
+        --enable-shared \
+        --enable-avresample \
+        --enable-libopencore-amrnb \
+        --enable-libopencore-amrwb \
+        --enable-gpl \
+        --enable-libass \
+        --enable-fontconfig \
+        --enable-libfreetype \
+        --enable-libvidstab \
+        --enable-libmp3lame \
+        --enable-libopus \
+        --enable-libtheora \
+        --enable-libvorbis \
+        --enable-libvpx \
+        --enable-libwebp \
+        --enable-libxcb \
+        --enable-libx265 \
+        --enable-libxvid \
+        --enable-libx264 \
+        --enable-nonfree \
+        --enable-openssl \
+        --enable-libfdk_aac \
+        --enable-postproc \
+        --enable-small \
+        --enable-version3 \
+        --enable-libbluray \
+        --enable-libzmq \
+        --extra-libs=-ldl \
+        --prefix="${PREFIX}" \
+        --enable-libopenjpeg \
+        --enable-libkvazaar \
+        --enable-libaom \
+        --extra-libs=-lpthread \
+        --enable-vaapi \
+        --extra-cflags="-I${PREFIX}/include" \
+        --extra-ldflags="-L${PREFIX}/lib" && \
+        make && \
+        make install && \
+        make tools/zmqsend && cp tools/zmqsend ${PREFIX}/bin/ && \
+        make distclean && \
+        hash -r && \
+        cd tools && \
+        make qt-faststart && cp qt-faststart ${PREFIX}/bin/
+
+## cleanup
+RUN \
+        ldd ${PREFIX}/bin/ffmpeg | grep opt/ffmpeg | cut -d ' ' -f 3 | xargs -i cp {} /usr/local/lib/ && \
+        for lib in /usr/local/lib/*.so.*; do ln -s "${lib##*/}" "${lib%%.so.*}".so; done && \
+        cp ${PREFIX}/bin/* /usr/local/bin/ && \
+        cp -r ${PREFIX}/share/ffmpeg /usr/local/share/ && \
+        LD_LIBRARY_PATH=/usr/local/lib ffmpeg -buildconf && \
+        cp -r ${PREFIX}/include/libav* ${PREFIX}/include/libpostproc ${PREFIX}/include/libsw* /usr/local/include && \
+        mkdir -p /usr/local/lib/pkgconfig && \
+        for pc in ${PREFIX}/lib/pkgconfig/libav*.pc ${PREFIX}/lib/pkgconfig/libpostproc.pc ${PREFIX}/lib/pkgconfig/libsw*.pc; do \
+          sed "s:${PREFIX}:/usr/local:g" <"$pc" >/usr/local/lib/pkgconfig/"${pc##*/}"; \
+        done
+
+FROM        base AS release
+
+ENV         LD_LIBRARY_PATH=/usr/local/lib:/usr/local/lib64:/usr/lib:/usr/lib64:/lib:/lib64
+
+CMD         ["--help"]
+ENTRYPOINT  ["ffmpeg"]
+
+COPY --from=build /usr/local /usr/local/
+
+RUN \
+	apt-get update -y && \
+	apt-get install -y --no-install-recommends libva-drm2 libva2 i965-va-driver && \
+	rm -rf /var/lib/apt/lists/*
--- a/docker/Dockerfile.ffmpeg.armv7
+++ b/docker/Dockerfile.ffmpeg.armv7
@@ -0,0 +1,549 @@
+# inspired by:
+# https://github.com/collelog/ffmpeg/blob/master/4.3.1-alpine-rpi4-arm64v8.Dockerfile
+# https://github.com/mmastrac/ffmpeg-omx-rpi-docker/blob/master/Dockerfile
+# https://github.com/jrottenberg/ffmpeg/pull/158/files
+# https://github.com/jrottenberg/ffmpeg/pull/239
+FROM        ubuntu:20.04 AS base
+
+WORKDIR     /tmp/workdir
+
+ENV     DEBIAN_FRONTEND=noninteractive
+RUN     apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends ca-certificates expat libgomp1 && \
+        apt-get autoremove -y && \
+        apt-get clean -y
+
+FROM base as build
+
+ENV         FFMPEG_VERSION=4.3.1 \
+            AOM_VERSION=v1.0.0 \
+            FDKAAC_VERSION=0.1.5 \
+            FONTCONFIG_VERSION=2.12.4 \
+            FREETYPE_VERSION=2.5.5 \
+            FRIBIDI_VERSION=0.19.7 \
+            KVAZAAR_VERSION=1.2.0 \
+            LAME_VERSION=3.100 \
+            LIBASS_VERSION=0.13.7 \
+            LIBPTHREAD_STUBS_VERSION=0.4 \
+            LIBVIDSTAB_VERSION=1.1.0 \
+            LIBXCB_VERSION=1.13.1 \
+            XCBPROTO_VERSION=1.13 \
+            OGG_VERSION=1.3.2 \
+            OPENCOREAMR_VERSION=0.1.5 \
+            OPUS_VERSION=1.2 \
+            OPENJPEG_VERSION=2.1.2 \
+            THEORA_VERSION=1.1.1 \
+            VORBIS_VERSION=1.3.5 \
+            VPX_VERSION=1.8.0 \
+            WEBP_VERSION=1.0.2 \
+            X264_VERSION=20170226-2245-stable \
+            X265_VERSION=3.1.1 \
+            XAU_VERSION=1.0.9 \
+            XORG_MACROS_VERSION=1.19.2 \
+            XPROTO_VERSION=7.0.31 \
+            XVID_VERSION=1.3.4 \
+            LIBXML2_VERSION=2.9.10 \
+            LIBBLURAY_VERSION=1.1.2 \
+            LIBZMQ_VERSION=4.3.3 \
+            SRC=/usr/local
+
+ARG         FREETYPE_SHA256SUM="5d03dd76c2171a7601e9ce10551d52d4471cf92cd205948e60289251daddffa8 freetype-2.5.5.tar.gz"
+ARG         FRIBIDI_SHA256SUM="3fc96fa9473bd31dcb5500bdf1aa78b337ba13eb8c301e7c28923fea982453a8 0.19.7.tar.gz"
+ARG         LIBASS_SHA256SUM="8fadf294bf701300d4605e6f1d92929304187fca4b8d8a47889315526adbafd7 0.13.7.tar.gz"
+ARG         LIBVIDSTAB_SHA256SUM="14d2a053e56edad4f397be0cb3ef8eb1ec3150404ce99a426c4eb641861dc0bb v1.1.0.tar.gz"
+ARG         OGG_SHA256SUM="e19ee34711d7af328cb26287f4137e70630e7261b17cbe3cd41011d73a654692 libogg-1.3.2.tar.gz"
+ARG         OPUS_SHA256SUM="77db45a87b51578fbc49555ef1b10926179861d854eb2613207dc79d9ec0a9a9 opus-1.2.tar.gz"
+ARG         THEORA_SHA256SUM="40952956c47811928d1e7922cda3bc1f427eb75680c3c37249c91e949054916b libtheora-1.1.1.tar.gz"
+ARG         VORBIS_SHA256SUM="6efbcecdd3e5dfbf090341b485da9d176eb250d893e3eb378c428a2db38301ce libvorbis-1.3.5.tar.gz"
+ARG         XVID_SHA256SUM="4e9fd62728885855bc5007fe1be58df42e5e274497591fec37249e1052ae316f xvidcore-1.3.4.tar.gz"
+ARG         LIBXML2_SHA256SUM="f07dab13bf42d2b8db80620cce7419b3b87827cc937c8bb20fe13b8571ee9501  libxml2-v2.9.10.tar.gz"
+ARG         LIBBLURAY_SHA256SUM="a3dd452239b100dc9da0d01b30e1692693e2a332a7d29917bf84bb10ea7c0b42 libbluray-1.1.2.tar.bz2"
+
+
+ARG         LD_LIBRARY_PATH=/opt/ffmpeg/lib
+ARG         MAKEFLAGS="-j2"
+ARG         PKG_CONFIG_PATH="/opt/ffmpeg/share/pkgconfig:/opt/ffmpeg/lib/pkgconfig:/opt/ffmpeg/lib64/pkgconfig:/opt/vc/lib/pkgconfig"
+ARG         PREFIX=/opt/ffmpeg
+ARG         LD_LIBRARY_PATH="/opt/ffmpeg/lib:/opt/ffmpeg/lib64:/usr/lib64:/usr/lib:/lib64:/lib:/opt/vc/lib"
+
+
+RUN      buildDeps="autoconf \
+                    automake \
+                    cmake \
+                    curl \
+                    bzip2 \
+                    libexpat1-dev \
+                    g++ \
+                    gcc \
+                    git \
+                    gperf \
+                    libtool \
+                    make \
+                    nasm \
+                    perl \
+                    pkg-config \
+                    python \
+                    sudo \
+                    libssl-dev \
+                    yasm \
+                    linux-headers-raspi2 \
+                    libomxil-bellagio-dev \
+                    libx265-dev \
+                    libaom-dev \
+                    zlib1g-dev" && \
+        apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends ${buildDeps}
+## opencore-amr https://sourceforge.net/projects/opencore-amr/
+RUN \
+        DIR=/tmp/opencore-amr && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://versaweb.dl.sourceforge.net/project/opencore-amr/opencore-amr/opencore-amr-${OPENCOREAMR_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## x264 http://www.videolan.org/developers/x264.html
+RUN \
+        DIR=/tmp/x264 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://download.videolan.org/pub/videolan/x264/snapshots/x264-snapshot-${X264_VERSION}.tar.bz2 | \
+        tar -jx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared --enable-pic --disable-cli && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+# ### x265 http://x265.org/
+# RUN \
+#         DIR=/tmp/x265 && \
+#         mkdir -p ${DIR} && \
+#         cd ${DIR} && \
+#         curl -sL https://download.videolan.org/pub/videolan/x265/x265_${X265_VERSION}.tar.gz  | \
+#         tar -zx && \
+#         cd x265_${X265_VERSION}/build/linux && \
+#         sed -i "/-DEXTRA_LIB/ s/$/ -DCMAKE_INSTALL_PREFIX=\${PREFIX}/" multilib.sh && \
+#         sed -i "/^cmake/ s/$/ -DENABLE_CLI=OFF/" multilib.sh && \
+#         # export CXXFLAGS="${CXXFLAGS} -fPIC" && \
+#         ./multilib.sh && \
+#         make -C 8bit install && \
+#         rm -rf ${DIR}
+### libogg https://www.xiph.org/ogg/
+RUN \
+        DIR=/tmp/ogg && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/ogg/libogg-${OGG_VERSION}.tar.gz && \
+        echo ${OGG_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libogg-${OGG_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libopus https://www.opus-codec.org/
+RUN \
+        DIR=/tmp/opus && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://archive.mozilla.org/pub/opus/opus-${OPUS_VERSION}.tar.gz && \
+        echo ${OPUS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f opus-${OPUS_VERSION}.tar.gz && \
+        autoreconf -fiv && \
+        ./configure --prefix="${PREFIX}" --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libvorbis https://xiph.org/vorbis/
+RUN \
+        DIR=/tmp/vorbis && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/vorbis/libvorbis-${VORBIS_VERSION}.tar.gz && \
+        echo ${VORBIS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libvorbis-${VORBIS_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --with-ogg="${PREFIX}" --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libtheora http://www.theora.org/
+RUN \
+        DIR=/tmp/theora && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xiph.org/releases/theora/libtheora-${THEORA_VERSION}.tar.gz && \
+        echo ${THEORA_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f libtheora-${THEORA_VERSION}.tar.gz && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD' -o config.guess && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD' -o config.sub && \
+        ./configure --prefix="${PREFIX}" --with-ogg="${PREFIX}" --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libvpx https://www.webmproject.org/code/
+RUN \
+        DIR=/tmp/vpx && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://codeload.github.com/webmproject/libvpx/tar.gz/v${VPX_VERSION} | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-vp8 --enable-vp9 --enable-vp9-highbitdepth --enable-pic --enable-shared \
+        --disable-debug --disable-examples --disable-docs --disable-install-bins  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libwebp https://developers.google.com/speed/webp/
+RUN \
+        DIR=/tmp/vebp && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-${WEBP_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --enable-shared  && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### libmp3lame http://lame.sourceforge.net/
+RUN \
+        DIR=/tmp/lame && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://versaweb.dl.sourceforge.net/project/lame/lame/$(echo ${LAME_VERSION} | sed -e 's/[^0-9]*\([0-9]*\)[.]\([0-9]*\)[.]\([0-9]*\)\([0-9A-Za-z-]*\)/\1.\2/')/lame-${LAME_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        ./configure --prefix="${PREFIX}" --bindir="${PREFIX}/bin" --enable-shared --enable-nasm --disable-frontend && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### xvid https://www.xvid.com/
+RUN \
+        DIR=/tmp/xvid && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO http://downloads.xvid.org/downloads/xvidcore-${XVID_VERSION}.tar.gz && \
+        echo ${XVID_SHA256SUM} | sha256sum --check && \
+        tar -zx -f xvidcore-${XVID_VERSION}.tar.gz && \
+        cd xvidcore/build/generic && \
+        ./configure --prefix="${PREFIX}" --bindir="${PREFIX}/bin" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+### fdk-aac https://github.com/mstorsjo/fdk-aac
+RUN \
+        DIR=/tmp/fdk-aac && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://github.com/mstorsjo/fdk-aac/archive/v${FDKAAC_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        autoreconf -fiv && \
+        ./configure --prefix="${PREFIX}" --enable-shared --datadir="${DIR}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## openjpeg https://github.com/uclouvain/openjpeg
+RUN \
+        DIR=/tmp/openjpeg && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sL https://github.com/uclouvain/openjpeg/archive/v${OPENJPEG_VERSION}.tar.gz | \
+        tar -zx --strip-components=1 && \
+        export CFLAGS="${CFLAGS} -DPNG_ARM_NEON_OPT=0" && \
+        cmake -DBUILD_THIRDPARTY:BOOL=ON -DCMAKE_INSTALL_PREFIX="${PREFIX}" . && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## freetype https://www.freetype.org/
+RUN  \
+        DIR=/tmp/freetype && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://download.savannah.gnu.org/releases/freetype/freetype-${FREETYPE_VERSION}.tar.gz && \
+        echo ${FREETYPE_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f freetype-${FREETYPE_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## libvstab https://github.com/georgmartius/vid.stab
+RUN  \
+        DIR=/tmp/vid.stab && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/georgmartius/vid.stab/archive/v${LIBVIDSTAB_VERSION}.tar.gz && \
+        echo ${LIBVIDSTAB_SHA256SUM} | sha256sum --check &&  \
+        tar -zx --strip-components=1 -f v${LIBVIDSTAB_VERSION}.tar.gz && \
+        cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" . && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## fridibi https://www.fribidi.org/
+RUN  \
+        DIR=/tmp/fribidi && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/fribidi/fribidi/archive/${FRIBIDI_VERSION}.tar.gz && \
+        echo ${FRIBIDI_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f ${FRIBIDI_VERSION}.tar.gz && \
+        sed -i 's/^SUBDIRS =.*/SUBDIRS=gen.tab charset lib bin/' Makefile.am && \
+        ./bootstrap --no-config --auto && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j1 && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## fontconfig https://www.freedesktop.org/wiki/Software/fontconfig/
+RUN  \
+        DIR=/tmp/fontconfig && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.freedesktop.org/software/fontconfig/release/fontconfig-${FONTCONFIG_VERSION}.tar.bz2 && \
+        tar -jx --strip-components=1 -f fontconfig-${FONTCONFIG_VERSION}.tar.bz2 && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## libass https://github.com/libass/libass
+RUN  \
+        DIR=/tmp/libass && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/libass/libass/archive/${LIBASS_VERSION}.tar.gz && \
+        echo ${LIBASS_SHA256SUM} | sha256sum --check && \
+        tar -zx --strip-components=1 -f ${LIBASS_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+## kvazaar https://github.com/ultravideo/kvazaar
+RUN \
+        DIR=/tmp/kvazaar && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/ultravideo/kvazaar/archive/v${KVAZAAR_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f v${KVAZAAR_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+# RUN \
+#         DIR=/tmp/aom && \
+#         git clone --branch ${AOM_VERSION} --depth 1 https://aomedia.googlesource.com/aom ${DIR} ; \
+#         cd ${DIR} ; \
+#         rm -rf CMakeCache.txt CMakeFiles ; \
+#         mkdir -p ./aom_build ; \
+#         cd ./aom_build ; \
+#         cmake -DCMAKE_INSTALL_PREFIX="${PREFIX}" -DBUILD_SHARED_LIBS=1 ..; \
+#         make ; \
+#         make install ; \
+#         rm -rf ${DIR}
+
+## libxcb (and supporting libraries) for screen capture https://xcb.freedesktop.org/
+RUN \
+        DIR=/tmp/xorg-macros && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive//individual/util/util-macros-${XORG_MACROS_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f util-macros-${XORG_MACROS_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/xproto && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive/individual/proto/xproto-${XPROTO_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f xproto-${XPROTO_VERSION}.tar.gz && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD' -o config.guess && \
+        curl -sL 'http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD' -o config.sub && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libXau && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://www.x.org/archive/individual/lib/libXau-${XAU_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libXau-${XAU_VERSION}.tar.gz && \
+        ./configure --srcdir=${DIR} --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libpthread-stubs && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/libpthread-stubs-${LIBPTHREAD_STUBS_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libpthread-stubs-${LIBPTHREAD_STUBS_VERSION}.tar.gz && \
+        ./configure --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libxcb-proto && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/xcb-proto-${XCBPROTO_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f xcb-proto-${XCBPROTO_VERSION}.tar.gz && \
+        ACLOCAL_PATH="${PREFIX}/share/aclocal" ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+RUN \
+        DIR=/tmp/libxcb && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://xcb.freedesktop.org/dist/libxcb-${LIBXCB_VERSION}.tar.gz && \
+        tar -zx --strip-components=1 -f libxcb-${LIBXCB_VERSION}.tar.gz && \
+        ACLOCAL_PATH="${PREFIX}/share/aclocal" ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## libxml2 - for libbluray
+RUN \
+        DIR=/tmp/libxml2 && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://gitlab.gnome.org/GNOME/libxml2/-/archive/v${LIBXML2_VERSION}/libxml2-v${LIBXML2_VERSION}.tar.gz && \
+        echo ${LIBXML2_SHA256SUM} | sha256sum --check && \
+        tar -xz --strip-components=1 -f libxml2-v${LIBXML2_VERSION}.tar.gz && \
+        ./autogen.sh --prefix="${PREFIX}" --with-ftp=no --with-http=no --with-python=no && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## libbluray - Requires libxml, freetype, and fontconfig
+RUN \
+        DIR=/tmp/libbluray && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://download.videolan.org/pub/videolan/libbluray/${LIBBLURAY_VERSION}/libbluray-${LIBBLURAY_VERSION}.tar.bz2 && \
+        echo ${LIBBLURAY_SHA256SUM} | sha256sum --check && \
+        tar -jx --strip-components=1 -f libbluray-${LIBBLURAY_VERSION}.tar.bz2 && \
+        ./configure --prefix="${PREFIX}" --disable-examples --disable-bdjava-jar --disable-static --enable-shared && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## libzmq https://github.com/zeromq/libzmq/
+RUN \
+        DIR=/tmp/libzmq && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        curl -sLO https://github.com/zeromq/libzmq/archive/v${LIBZMQ_VERSION}.tar.gz && \
+        tar -xz --strip-components=1 -f v${LIBZMQ_VERSION}.tar.gz && \
+        ./autogen.sh && \
+        ./configure --prefix="${PREFIX}" && \
+        make -j $(nproc) && \
+        # make check && \
+        make -j $(nproc) install && \
+        rm -rf ${DIR}
+
+## userland https://github.com/raspberrypi/userland
+RUN \
+        DIR=/tmp/userland && \
+        mkdir -p ${DIR} && \
+        cd ${DIR} && \
+        git clone --depth 1 https://github.com/raspberrypi/userland.git . && \
+        ./buildme && \
+        rm -rf ${DIR}
+
+## ffmpeg https://ffmpeg.org/
+RUN  \
+        DIR=/tmp/ffmpeg && mkdir -p ${DIR} && cd ${DIR} && \
+        curl -sLO https://ffmpeg.org/releases/ffmpeg-${FFMPEG_VERSION}.tar.bz2 && \
+        tar -jx --strip-components=1 -f ffmpeg-${FFMPEG_VERSION}.tar.bz2
+
+RUN \
+        DIR=/tmp/ffmpeg && mkdir -p ${DIR} && cd ${DIR} && \
+        ./configure \
+        --disable-debug \
+        --disable-doc \
+        --disable-ffplay \
+        --enable-shared \
+        --enable-avresample \
+        --enable-libopencore-amrnb \
+        --enable-libopencore-amrwb \
+        --enable-gpl \
+        --enable-libass \
+        --enable-fontconfig \
+        --enable-libfreetype \
+        --enable-libvidstab \
+        --enable-libmp3lame \
+        --enable-libopus \
+        --enable-libtheora \
+        --enable-libvorbis \
+        --enable-libvpx \
+        --enable-libwebp \
+        --enable-libxcb \
+        --enable-libx265 \
+        --enable-libxvid \
+        --enable-libx264 \
+        --enable-nonfree \
+        --enable-openssl \
+        --enable-libfdk_aac \
+        --enable-postproc \
+        --enable-small \
+        --enable-version3 \
+        --enable-libbluray \
+        --enable-libzmq \
+        --extra-libs=-ldl \
+        --prefix="${PREFIX}" \
+        --enable-libopenjpeg \
+        --enable-libkvazaar \
+        --enable-libaom \
+        --extra-libs=-lpthread \
+        --enable-omx \
+        --enable-omx-rpi \
+        --enable-mmal \
+        --enable-v4l2_m2m \
+        --enable-neon \
+        --extra-cflags="-I${PREFIX}/include" \
+        --extra-ldflags="-L${PREFIX}/lib" && \
+        make -j $(nproc) && \
+        make -j $(nproc) install && \
+        make tools/zmqsend && cp tools/zmqsend ${PREFIX}/bin/ && \
+        make distclean && \
+        hash -r && \
+        cd tools && \
+        make qt-faststart && cp qt-faststart ${PREFIX}/bin/
+
+## cleanup
+RUN \
+        ldd ${PREFIX}/bin/ffmpeg | grep opt/ffmpeg | cut -d ' ' -f 3 | xargs -i cp {} /usr/local/lib/ && \
+        # copy userland lib too
+        ldd ${PREFIX}/bin/ffmpeg | grep opt/vc | cut -d ' ' -f 3 | xargs -i cp {} /usr/local/lib/ && \
+        for lib in /usr/local/lib/*.so.*; do ln -s "${lib##*/}" "${lib%%.so.*}".so; done && \
+        cp ${PREFIX}/bin/* /usr/local/bin/ && \
+        cp -r ${PREFIX}/share/ffmpeg /usr/local/share/ && \
+        LD_LIBRARY_PATH=/usr/local/lib ffmpeg -buildconf && \
+        cp -r ${PREFIX}/include/libav* ${PREFIX}/include/libpostproc ${PREFIX}/include/libsw* /usr/local/include && \
+        mkdir -p /usr/local/lib/pkgconfig && \
+        for pc in ${PREFIX}/lib/pkgconfig/libav*.pc ${PREFIX}/lib/pkgconfig/libpostproc.pc ${PREFIX}/lib/pkgconfig/libsw*.pc; do \
+          sed "s:${PREFIX}:/usr/local:g" <"$pc" >/usr/local/lib/pkgconfig/"${pc##*/}"; \
+        done
+
+FROM        base AS release
+
+ENV         LD_LIBRARY_PATH=/usr/local/lib:/usr/local/lib64:/usr/lib:/usr/lib64:/lib:/lib64
+
+RUN \
+        apt-get -yqq update && \
+        apt-get install -yq --no-install-recommends libx265-dev libaom-dev && \
+        apt-get autoremove -y && \
+        apt-get clean -y
+
+CMD         ["--help"]
+ENTRYPOINT  ["ffmpeg"]
+
+COPY --from=build /usr/local /usr/local/
--- a/docker/Dockerfile.wheels
+++ b/docker/Dockerfile.wheels
@@ -0,0 +1,39 @@
+FROM ubuntu:20.04 as build
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get -qq update \
+    && apt-get -qq install -y \
+    python3 \
+    python3-dev \
+    wget \
+    # opencv dependencies
+    build-essential cmake git pkg-config libgtk-3-dev \
+    libavcodec-dev libavformat-dev libswscale-dev libv4l-dev \
+    libxvidcore-dev libx264-dev libjpeg-dev libpng-dev libtiff-dev \
+    gfortran openexr libatlas-base-dev libssl-dev\
+    libtbb2 libtbb-dev libdc1394-22-dev libopenexr-dev \
+    libgstreamer-plugins-base1.0-dev libgstreamer1.0-dev \
+    # scipy dependencies
+    gcc gfortran libopenblas-dev liblapack-dev cython
+
+RUN wget -q https://bootstrap.pypa.io/get-pip.py -O get-pip.py \
+    && python3 get-pip.py
+
+RUN pip3 install scikit-build
+
+RUN pip3 wheel --wheel-dir=/wheels \
+    opencv-python-headless \
+    numpy \
+    imutils \
+    scipy \
+    psutil \
+    Flask \
+    paho-mqtt \
+    PyYAML \
+    matplotlib \
+    click
+
+FROM scratch
+
+COPY --from=build /wheels /wheels
--- a/docker/Dockerfile.wheels.aarch64
+++ b/docker/Dockerfile.wheels.aarch64
@@ -0,0 +1,49 @@
+FROM ubuntu:20.04 as build
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get -qq update \
+    && apt-get -qq install -y \
+    python3 \
+    python3-dev \
+    wget \
+    # opencv dependencies
+    build-essential cmake git pkg-config libgtk-3-dev \
+    libavcodec-dev libavformat-dev libswscale-dev libv4l-dev \
+    libxvidcore-dev libx264-dev libjpeg-dev libpng-dev libtiff-dev \
+    gfortran openexr libatlas-base-dev libssl-dev\
+    libtbb2 libtbb-dev libdc1394-22-dev libopenexr-dev \
+    libgstreamer-plugins-base1.0-dev libgstreamer1.0-dev \
+    # scipy dependencies
+    gcc gfortran libopenblas-dev liblapack-dev cython
+
+RUN wget -q https://bootstrap.pypa.io/get-pip.py -O get-pip.py \
+    && python3 get-pip.py
+
+# need to build cmake from source because binary distribution is broken for arm64
+# https://github.com/scikit-build/cmake-python-distributions/issues/115
+# https://github.com/skvark/opencv-python/issues/366
+# https://github.com/scikit-build/cmake-python-distributions/issues/96#issuecomment-663062358
+RUN pip3 install scikit-build
+
+RUN git clone https://github.com/scikit-build/cmake-python-distributions.git \
+    && cd cmake-python-distributions/ \
+    && python3 setup.py bdist_wheel
+
+RUN pip3 install cmake-python-distributions/dist/*.whl
+
+RUN pip3 wheel --wheel-dir=/wheels \
+    opencv-python-headless \
+    numpy \
+    imutils \
+    scipy \
+    psutil \
+    Flask \
+    paho-mqtt \
+    PyYAML \
+    matplotlib \
+    click
+
+FROM scratch
+
+COPY --from=build /wheels /wheels
--- a/docs/DEVICES.md
+++ b/docs/DEVICES.md
@@ -1,74 +0,0 @@
-# Configuration Examples
-
-### Default (most RTSP cameras)
-This is the default ffmpeg command and should work with most RTSP cameras that send h264 video
-```yaml
-ffmpeg:
-  global_args:
-    - -hide_banner
-    - -loglevel
-    - panic
-  hwaccel_args: []
-  input_args:
-    - -avoid_negative_ts
-    - make_zero 
-    - -fflags
-    - nobuffer
-    - -flags
-    - low_delay
-    - -strict
-    - experimental
-    - -fflags
-    - +genpts+discardcorrupt
-    - -vsync
-    - drop
-    - -rtsp_transport
-    - tcp
-    - -stimeout
-    - '5000000' 
-    - -use_wallclock_as_timestamps
-    - '1'
-  output_args:
-    - -vf
-    - mpdecimate
-    - -f
-    - rawvideo
-    - -pix_fmt
-    - rgb24
-```
-
-### RTMP Cameras
-The input parameters need to be adjusted for RTMP cameras
-```yaml
-ffmpeg:
-input_args:
-    - -avoid_negative_ts
-    - make_zero
-    - -fflags
-    - nobuffer
-    - -flags
-    - low_delay
-    - -strict
-    - experimental
-    - -fflags
-    - +genpts+discardcorrupt
-    - -vsync
-    - drop
-    - -use_wallclock_as_timestamps
-    - '1'
-```
-
-
-### Hardware Acceleration
-
-Intel Quicksync
-```yaml
-ffmpeg:
-  hwaccel_args:
-    - -hwaccel
-    - vaapi
-    - -hwaccel_device
-    - /dev/dri/renderD128
-    - -hwaccel_output_format
-    - yuv420p
-```
--- a/docs/cameras.md
+++ b/docs/cameras.md
@@ -0,0 +1,21 @@
+# Camera Specific Configuration
+Frigate should work with most RTSP cameras and h264 feeds such as Dahua.
+
+## RTMP Cameras
+The input parameters need to be adjusted for RTMP cameras
+```yaml
+ffmpeg:
+input_args:
+    - -avoid_negative_ts
+    - make_zero
+    - -fflags
+    - nobuffer
+    - -flags
+    - low_delay
+    - -strict
+    - experimental
+    - -fflags
+    - +genpts+discardcorrupt
+    - -use_wallclock_as_timestamps
+    - '1'
+```
--- a/docs/diagram.png
+++ b/docs/diagram.png
--- a/docs/example-mask-poly.png
+++ b/docs/example-mask-poly.png
--- a/docs/frigate.png
+++ b/docs/frigate.png
--- a/docs/how-frigate-works.md
+++ b/docs/how-frigate-works.md
@@ -0,0 +1,10 @@
+# How Frigate Works
+Frigate is designed to minimize resource and maximize performance by only looking for objects when and where it is necessary
+
+![Diagram](diagram.png)
+
+## 1. Look for Motion
+
+## 2. Calculate Detection Regions
+
+## 3. Run Object Detection
--- a/frigate/edgetpu.py
+++ b/frigate/edgetpu.py
@@ -2,12 +2,14 @@ import os
 import datetime
 import hashlib
 import multiprocessing as mp
+import queue
+from multiprocessing.connection import Connection
 from abc import ABC, abstractmethod
+from typing import Dict
 import numpy as np
-import pyarrow.plasma as plasma
 import tflite_runtime.interpreter as tflite
 from tflite_runtime.interpreter import load_delegate
-from frigate.util import EventsPerSecond, listen
+from frigate.util import EventsPerSecond, listen, SharedMemoryFrameManager

 def load_labels(path, encoding='utf-8'):
  """Loads labels from file (with or without index numbers).
@@ -46,15 +48,12 @@ class LocalObjectDetector(ObjectDetector):
            device_config = {"device": tf_device}

        edge_tpu_delegate = None
-        try:
-            print(f"Attempting to load TPU as {device_config['device']}")
-            edge_tpu_delegate = load_delegate('libedgetpu.so.1.0', device_config)
-            print("TPU found")
-        except ValueError:
+
+        if tf_device != 'cpu':
            try:
-                print(f"Attempting to load TPU as pci:0")
-                edge_tpu_delegate = load_delegate('libedgetpu.so.1.0', {"device": "pci:0"})
-                print("PCIe TPU found")
+                print(f"Attempting to load TPU as {device_config['device']}")
+                edge_tpu_delegate = load_delegate('libedgetpu.so.1.0', device_config)
+                print("TPU found")
            except ValueError:
                print("No EdgeTPU detected. Falling back to CPU.")
        
@@ -100,76 +99,91 @@ class LocalObjectDetector(ObjectDetector):
        
        return detections

-def run_detector(detection_queue, avg_speed, start, tf_device):
+def run_detector(detection_queue, out_events: Dict[str, mp.Event], avg_speed, start, tf_device):
    print(f"Starting detection process: {os.getpid()}")
    listen()
-    plasma_client = plasma.connect("/tmp/plasma")
+    frame_manager = SharedMemoryFrameManager()
    object_detector = LocalObjectDetector(tf_device=tf_device)

+    outputs = {}
+    for name in out_events.keys():
+        out_shm = mp.shared_memory.SharedMemory(name=f"out-{name}", create=False)
+        out_np = np.ndarray((20,6), dtype=np.float32, buffer=out_shm.buf)
+        outputs[name] = {
+            'shm': out_shm,
+            'np': out_np
+        }
+    
    while True:
-        object_id_str = detection_queue.get()
-        object_id_hash = hashlib.sha1(str.encode(object_id_str))
-        object_id = plasma.ObjectID(object_id_hash.digest())
-        object_id_out = plasma.ObjectID(hashlib.sha1(str.encode(f"out-{object_id_str}")).digest())
-        input_frame = plasma_client.get(object_id, timeout_ms=0)
+        connection_id = detection_queue.get()
+        input_frame = frame_manager.get(connection_id, (1,300,300,3))

-        if input_frame is plasma.ObjectNotAvailable:
+        if input_frame is None:
            continue

-        # detect and put the output in the plasma store
+        # detect and send the output
        start.value = datetime.datetime.now().timestamp()
-        plasma_client.put(object_detector.detect_raw(input_frame), object_id_out)
+        detections = object_detector.detect_raw(input_frame)
        duration = datetime.datetime.now().timestamp()-start.value
+        outputs[connection_id]['np'][:] = detections[:]
+        out_events[connection_id].set()
        start.value = 0.0

        avg_speed.value = (avg_speed.value*9 + duration)/10
        
 class EdgeTPUProcess():
-    def __init__(self, tf_device=None):
-        self.detection_queue = mp.Queue()
+    def __init__(self, detection_queue, out_events, tf_device=None):
+        self.out_events = out_events
+        self.detection_queue = detection_queue
        self.avg_inference_speed = mp.Value('d', 0.01)
        self.detection_start = mp.Value('d', 0.0)
        self.detect_process = None
        self.tf_device = tf_device
        self.start_or_restart()
+    
+    def stop(self):
+        self.detect_process.terminate()
+        print("Waiting for detection process to exit gracefully...")
+        self.detect_process.join(timeout=30)
+        if self.detect_process.exitcode is None:
+            print("Detection process didnt exit. Force killing...")
+            self.detect_process.kill()
+            self.detect_process.join()

    def start_or_restart(self):
        self.detection_start.value = 0.0
        if (not self.detect_process is None) and self.detect_process.is_alive():
-            self.detect_process.terminate()
-            print("Waiting for detection process to exit gracefully...")
-            self.detect_process.join(timeout=30)
-            if self.detect_process.exitcode is None:
-                print("Detection process didnt exit. Force killing...")
-                self.detect_process.kill()
-                self.detect_process.join()
-        self.detect_process = mp.Process(target=run_detector, args=(self.detection_queue, self.avg_inference_speed, self.detection_start, self.tf_device))
+            self.stop()
+        self.detect_process = mp.Process(target=run_detector, args=(self.detection_queue, self.out_events, self.avg_inference_speed, self.detection_start, self.tf_device))
        self.detect_process.daemon = True
        self.detect_process.start()

 class RemoteObjectDetector():
-    def __init__(self, name, labels, detection_queue):
+    def __init__(self, name, labels, detection_queue, event):
        self.labels = load_labels(labels)
        self.name = name
        self.fps = EventsPerSecond()
-        self.plasma_client = plasma.connect("/tmp/plasma")
        self.detection_queue = detection_queue
+        self.event = event
+        self.shm = mp.shared_memory.SharedMemory(name=self.name, create=False)
+        self.np_shm = np.ndarray((1,300,300,3), dtype=np.uint8, buffer=self.shm.buf)
+        self.out_shm = mp.shared_memory.SharedMemory(name=f"out-{self.name}", create=False)
+        self.out_np_shm = np.ndarray((20,6), dtype=np.float32, buffer=self.out_shm.buf)
    
    def detect(self, tensor_input, threshold=.4):
        detections = []

-        now = f"{self.name}-{str(datetime.datetime.now().timestamp())}"
-        object_id_frame = plasma.ObjectID(hashlib.sha1(str.encode(now)).digest())
-        object_id_detections = plasma.ObjectID(hashlib.sha1(str.encode(f"out-{now}")).digest())
-        self.plasma_client.put(tensor_input, object_id_frame)
-        self.detection_queue.put(now)
-        raw_detections = self.plasma_client.get(object_id_detections, timeout_ms=10000)
+        # copy input to shared memory
+        self.np_shm[:] = tensor_input[:]
+        self.event.clear()
+        self.detection_queue.put(self.name)
+        result = self.event.wait(timeout=10.0)

-        if raw_detections is plasma.ObjectNotAvailable:
-            self.plasma_client.delete([object_id_frame])
+        # if it timed out
+        if result is None:
            return detections

-        for d in raw_detections:
+        for d in self.out_np_shm:
            if d[1] < threshold:
                break
            detections.append((
@@ -177,6 +191,9 @@ class RemoteObjectDetector():
                float(d[1]),
                (d[2], d[3], d[4], d[5])
            ))
-        self.plasma_client.delete([object_id_frame, object_id_detections])
        self.fps.update()
        return detections
+    
+    def cleanup(self):
+        self.shm.unlink()
+        self.out_shm.unlink()
--- a/frigate/motion.py
+++ b/frigate/motion.py
@@ -4,6 +4,7 @@ import numpy as np

 class MotionDetector():
    def __init__(self, frame_shape, mask, resize_factor=4):
+        self.frame_shape = frame_shape
        self.resize_factor = resize_factor
        self.motion_frame_size = (int(frame_shape[0]/resize_factor), int(frame_shape[1]/resize_factor))
        self.avg_frame = np.zeros(self.motion_frame_size, np.float)
@@ -16,14 +17,16 @@ class MotionDetector():
    def detect(self, frame):
        motion_boxes = []

+        gray = frame[0:self.frame_shape[0], 0:self.frame_shape[1]]
+
        # resize frame
-        resized_frame = cv2.resize(frame, dsize=(self.motion_frame_size[1], self.motion_frame_size[0]), interpolation=cv2.INTER_LINEAR)
+        resized_frame = cv2.resize(gray, dsize=(self.motion_frame_size[1], self.motion_frame_size[0]), interpolation=cv2.INTER_LINEAR)

        # convert to grayscale
-        gray = cv2.cvtColor(resized_frame, cv2.COLOR_BGR2GRAY)
+        # resized_frame = cv2.cvtColor(resized_frame, cv2.COLOR_BGR2GRAY)

        # mask frame
-        gray[self.mask] = [255]
+        resized_frame[self.mask] = [255]

        # it takes ~30 frames to establish a baseline
        # dont bother looking for motion
@@ -31,7 +34,7 @@ class MotionDetector():
            self.frame_counter += 1
        else:
            # compare to average
-            frameDelta = cv2.absdiff(gray, cv2.convertScaleAbs(self.avg_frame))
+            frameDelta = cv2.absdiff(resized_frame, cv2.convertScaleAbs(self.avg_frame))

            # compute the average delta over the past few frames
            # the alpha value can be modified to configure how sensitive the motion detection is.
@@ -70,10 +73,10 @@ class MotionDetector():
            # TODO: this really depends on FPS
            if self.motion_frame_count >= 10:
                # only average in the current frame if the difference persists for at least 3 frames
-                cv2.accumulateWeighted(gray, self.avg_frame, 0.2)
+                cv2.accumulateWeighted(resized_frame, self.avg_frame, 0.2)
        else:
            # when no motion, just keep averaging the frames together
-            cv2.accumulateWeighted(gray, self.avg_frame, 0.2)
+            cv2.accumulateWeighted(resized_frame, self.avg_frame, 0.2)
            self.motion_frame_count = 0

        return motion_boxes
--- a/frigate/object_processing.py
+++ b/frigate/object_processing.py
@@ -10,9 +10,8 @@ import copy
 import numpy as np
 from collections import Counter, defaultdict
 import itertools
-import pyarrow.plasma as plasma
 import matplotlib.pyplot as plt
-from frigate.util import draw_box_with_label, PlasmaFrameManager
+from frigate.util import draw_box_with_label, SharedMemoryFrameManager
 from frigate.edgetpu import load_labels
 from typing import Callable, Dict
 from statistics import mean, median
@@ -59,11 +58,48 @@ class CameraState():
        self.object_status = defaultdict(lambda: 'OFF')
        self.tracked_objects = {}
        self.zone_objects = defaultdict(lambda: [])
-        self.current_frame = np.zeros((720,1280,3), np.uint8)
+        self._current_frame = np.zeros((self.config['frame_shape'][0]*3//2, self.config['frame_shape'][1]), np.uint8)
+        self.current_frame_lock = threading.Lock()
        self.current_frame_time = 0.0
        self.previous_frame_id = None
        self.callbacks = defaultdict(lambda: [])

+    def get_current_frame(self, draw=False):
+        with self.current_frame_lock:
+            frame_copy = np.copy(self._current_frame)
+            frame_time = self.current_frame_time
+            tracked_objects = copy.deepcopy(self.tracked_objects)
+        
+        frame_copy = cv2.cvtColor(frame_copy, cv2.COLOR_YUV2BGR_I420)
+        # draw on the frame
+        if draw:
+            # draw the bounding boxes on the frame
+            for obj in tracked_objects.values():
+                thickness = 2
+                color = COLOR_MAP[obj['label']]
+                
+                if obj['frame_time'] != frame_time:
+                    thickness = 1
+                    color = (255,0,0)
+
+                # draw the bounding boxes on the frame
+                box = obj['box']
+                draw_box_with_label(frame_copy, box[0], box[1], box[2], box[3], obj['label'], f"{int(obj['score']*100)}% {int(obj['area'])}", thickness=thickness, color=color)
+                # draw the regions on the frame
+                region = obj['region']
+                cv2.rectangle(frame_copy, (region[0], region[1]), (region[2], region[3]), (0,255,0), 1)
+            
+            if self.config['snapshots']['show_timestamp']:
+                time_to_show = datetime.datetime.fromtimestamp(frame_time).strftime("%m/%d/%Y %H:%M:%S")
+                cv2.putText(frame_copy, time_to_show, (10, 30), cv2.FONT_HERSHEY_SIMPLEX, fontScale=.8, color=(255, 255, 255), thickness=2)
+
+            if self.config['snapshots']['draw_zones']:
+                for name, zone in self.config['zones'].items():
+                    thickness = 8 if any([name in obj['zones'] for obj in tracked_objects.values()]) else 2
+                    cv2.drawContours(frame_copy, [zone['contour']], -1, zone['color'], thickness)
+        
+        return frame_copy
+
    def false_positive(self, obj):
        # once a true positive, always a true positive
        if not obj.get('false_positive', True):
@@ -88,10 +124,7 @@ class CameraState():
        self.current_frame_time = frame_time
        # get the new frame and delete the old frame
        frame_id = f"{self.name}{frame_time}"
-        self.current_frame = self.frame_manager.get(frame_id)
-        if not self.previous_frame_id is None:
-            self.frame_manager.delete(self.previous_frame_id)
-        self.previous_frame_id = frame_id
+        current_frame = self.frame_manager.get(frame_id, (self.config['frame_shape'][0]*3//2, self.config['frame_shape'][1]))

        current_ids = tracked_objects.keys()
        previous_ids = self.tracked_objects.keys()
@@ -148,38 +181,13 @@ class CameraState():
            # check each zone
            for name, zone in self.config['zones'].items():
                contour = zone['contour']
-                # check if the object is in the zone and not filtered
-                if (cv2.pointPolygonTest(contour, bottom_center, False) >= 0 
-                    and not zone_filtered(obj, zone.get('filters', {}))):
-                    current_zones.append(name)
+                # check if the object is in the zone
+                if (cv2.pointPolygonTest(contour, bottom_center, False) >= 0):
+                    # if the object passed the filters once, dont apply again
+                    if name in obj.get('zones', []) or not zone_filtered(obj, zone.get('filters', {})):
+                        current_zones.append(name)
+                    
            obj['zones'] = current_zones
-        
-        # draw on the frame
-        if not self.current_frame is None:
-            # draw the bounding boxes on the frame
-            for obj in self.tracked_objects.values():
-                thickness = 2
-                color = COLOR_MAP[obj['label']]
-                
-                if obj['frame_time'] != frame_time:
-                    thickness = 1
-                    color = (255,0,0)
-
-                # draw the bounding boxes on the frame
-                box = obj['box']
-                draw_box_with_label(self.current_frame, box[0], box[1], box[2], box[3], obj['label'], f"{int(obj['score']*100)}% {int(obj['area'])}", thickness=thickness, color=color)
-                # draw the regions on the frame
-                region = obj['region']
-                cv2.rectangle(self.current_frame, (region[0], region[1]), (region[2], region[3]), (0,255,0), 1)
-            
-            if self.config['snapshots']['show_timestamp']:
-                time_to_show = datetime.datetime.fromtimestamp(frame_time).strftime("%m/%d/%Y %H:%M:%S")
-                cv2.putText(self.current_frame, time_to_show, (10, 30), cv2.FONT_HERSHEY_SIMPLEX, fontScale=.8, color=(255, 255, 255), thickness=2)
-
-            if self.config['snapshots']['draw_zones']:
-                for name, zone in self.config['zones'].items():
-                    thickness = 8 if any([name in obj['zones'] for obj in self.tracked_objects.values()]) else 2
-                    cv2.drawContours(self.current_frame, [zone['contour']], -1, zone['color'], thickness)

        # maintain best objects
        for obj in self.tracked_objects.values():
@@ -194,12 +202,12 @@ class CameraState():
                # if the object is a higher score than the current best score 
                # or the current object is older than desired, use the new object
                if obj_copy['score'] > current_best['score'] or (now - current_best['frame_time']) > self.config.get('best_image_timeout', 60):
-                    obj_copy['frame'] = np.copy(self.current_frame)
+                    obj_copy['frame'] = np.copy(current_frame)
                    self.best_objects[object_type] = obj_copy
                    for c in self.callbacks['snapshot']:
                        c(self.name, self.best_objects[object_type])
            else:
-                obj_copy['frame'] = np.copy(self.current_frame)
+                obj_copy['frame'] = np.copy(current_frame)
                self.best_objects[object_type] = obj_copy
                for c in self.callbacks['snapshot']:
                    c(self.name, self.best_objects[object_type])
@@ -226,7 +234,12 @@ class CameraState():
                c(self.name, obj_name, 'OFF')
            for c in self.callbacks['snapshot']:
                c(self.name, self.best_objects[obj_name])
-
+        
+        with self.current_frame_lock:
+            self._current_frame = current_frame
+            if not self.previous_frame_id is None:
+                self.frame_manager.delete(self.previous_frame_id)
+            self.previous_frame_id = frame_id

 class TrackedObjectProcessor(threading.Thread):
    def __init__(self, camera_config, client, topic_prefix, tracked_objects_queue, event_queue, stop_event):
@@ -238,7 +251,7 @@ class TrackedObjectProcessor(threading.Thread):
        self.event_queue = event_queue
        self.stop_event = stop_event
        self.camera_states: Dict[str, CameraState] = {}
-        self.plasma_client = PlasmaFrameManager(self.stop_event)
+        self.frame_manager = SharedMemoryFrameManager()

        def start(camera, obj):
            # publish events to mqtt
@@ -255,7 +268,14 @@ class TrackedObjectProcessor(threading.Thread):
        def snapshot(camera, obj):
            if not 'frame' in obj:
                return
-            best_frame = cv2.cvtColor(obj['frame'], cv2.COLOR_RGB2BGR)
+            
+            best_frame = cv2.cvtColor(obj['frame'], cv2.COLOR_YUV2BGR_I420)
+            if self.camera_config[camera]['snapshots']['draw_bounding_boxes']:
+                thickness = 2
+                color = COLOR_MAP[obj['label']]
+                box = obj['box']
+                draw_box_with_label(best_frame, box[0], box[1], box[2], box[3], obj['label'], f"{int(obj['score']*100)}% {int(obj['area'])}", thickness=thickness, color=color)
+                
            mqtt_config = self.camera_config[camera].get('mqtt', {'crop_to_region': False})
            if mqtt_config.get('crop_to_region'):
                region = obj['region']
@@ -264,6 +284,16 @@ class TrackedObjectProcessor(threading.Thread):
                height = int(mqtt_config['snapshot_height'])
                width = int(height*best_frame.shape[1]/best_frame.shape[0])
                best_frame = cv2.resize(best_frame, dsize=(width, height), interpolation=cv2.INTER_AREA)
+            
+            if self.camera_config[camera]['snapshots']['show_timestamp']:
+                time_to_show = datetime.datetime.fromtimestamp(obj['frame_time']).strftime("%m/%d/%Y %H:%M:%S")
+                size = cv2.getTextSize(time_to_show, cv2.FONT_HERSHEY_SIMPLEX, fontScale=1, thickness=2)
+                text_width = size[0][0]
+                text_height = size[0][1]
+                desired_size = max(200, 0.33*best_frame.shape[1])
+                font_scale = desired_size/text_width
+                cv2.putText(best_frame, time_to_show, (5, best_frame.shape[0]-7), cv2.FONT_HERSHEY_SIMPLEX, fontScale=font_scale, color=(255, 255, 255), thickness=2)
+
            ret, jpg = cv2.imencode('.jpg', best_frame)
            if ret:
                jpg_bytes = jpg.tobytes()
@@ -273,7 +303,7 @@ class TrackedObjectProcessor(threading.Thread):
            self.client.publish(f"{self.topic_prefix}/{camera}/{object_name}", status, retain=False)

        for camera in self.camera_config.keys():
-            camera_state = CameraState(camera, self.camera_config[camera], self.plasma_client)
+            camera_state = CameraState(camera, self.camera_config[camera], self.frame_manager)
            camera_state.on('start', start)
            camera_state.on('update', update)
            camera_state.on('end', end)
@@ -323,8 +353,8 @@ class TrackedObjectProcessor(threading.Thread):
        else:
            return {}
    
-    def get_current_frame(self, camera):
-        return self.camera_states[camera].current_frame
+    def get_current_frame(self, camera, draw=False):
+        return self.camera_states[camera].get_current_frame(draw)

    def run(self):
        while True:
--- a/frigate/util.py
+++ b/frigate/util.py
@@ -9,7 +9,8 @@ import cv2
 import threading
 import matplotlib.pyplot as plt
 import hashlib
-import pyarrow.plasma as plasma
+from multiprocessing import shared_memory
+from typing import AnyStr

 def draw_box_with_label(frame, x_min, y_min, x_max, y_max, label, info, thickness=2, color=None, position='ul'):
    if color is None:
@@ -69,6 +70,37 @@ def calculate_region(frame_shape, xmin, ymin, xmax, ymax, multiplier=2):

    return (x_offset, y_offset, x_offset+size, y_offset+size)

+def yuv_region_2_rgb(frame, region):
+    height = frame.shape[0]//3*2
+    width = frame.shape[1]
+    # make sure the size is a multiple of 4
+    size = (region[3] - region[1])//4*4
+
+    x1 = region[0] 
+    y1 = region[1]
+
+    uv_x1 = x1//2
+    uv_y1 = y1//4
+
+    uv_width = size//2
+    uv_height = size//4
+
+    u_y_start = height
+    v_y_start = height + height//4
+    two_x_offset = width//2
+
+    yuv_cropped_frame = np.zeros((size+size//2, size), np.uint8)
+    # y channel
+    yuv_cropped_frame[0:size, 0:size] = frame[y1:y1+size, x1:x1+size]
+    # u channel
+    yuv_cropped_frame[size:size+uv_height, 0:uv_width] = frame[uv_y1+u_y_start:uv_y1+u_y_start+uv_height, uv_x1:uv_x1+uv_width]
+    yuv_cropped_frame[size:size+uv_height, uv_width:size] = frame[uv_y1+u_y_start:uv_y1+u_y_start+uv_height, uv_x1+two_x_offset:uv_x1+two_x_offset+uv_width]
+    # v channel
+    yuv_cropped_frame[size+uv_height:size+uv_height*2, 0:uv_width] = frame[uv_y1+v_y_start:uv_y1+v_y_start+uv_height, uv_x1:uv_x1+uv_width]
+    yuv_cropped_frame[size+uv_height:size+uv_height*2, uv_width:size] = frame[uv_y1+v_y_start:uv_y1+v_y_start+uv_height, uv_x1+two_x_offset:uv_x1+two_x_offset+uv_width]
+
+    return cv2.cvtColor(yuv_cropped_frame, cv2.COLOR_YUV2RGB_I420)
+
 def intersection(box_a, box_b):
    return (
        max(box_a[0], box_b[0]),
@@ -148,12 +180,16 @@ def listen():
    signal.signal(signal.SIGUSR1, print_stack)

 class FrameManager(ABC):
+    @abstractmethod
+    def create(self, name, size) -> AnyStr:
+        pass
+
    @abstractmethod
    def get(self, name, timeout_ms=0):
        pass

    @abstractmethod
-    def put(self, name, frame):
+    def close(self, name):
        pass

    @abstractmethod
@@ -164,66 +200,45 @@ class DictFrameManager(FrameManager):
    def __init__(self):
        self.frames = {}
    
-    def get(self, name, timeout_ms=0):
-        return self.frames.get(name)
+    def create(self, name, size) -> AnyStr:
+        mem = bytearray(size)
+        self.frames[name] = mem
+        return mem
    
-    def put(self, name, frame):
-        self.frames[name] = frame
+    def get(self, name, shape):
+        mem = self.frames[name]
+        return np.ndarray(shape, dtype=np.uint8, buffer=mem)
+    
+    def close(self, name):
+        pass
    
    def delete(self, name):
        del self.frames[name]

-class PlasmaFrameManager(FrameManager):
-    def __init__(self, stop_event=None):
-        self.stop_event = stop_event
-        self.connect()
+class SharedMemoryFrameManager(FrameManager):
+    def __init__(self):
+        self.shm_store = {}
    
-    def connect(self):
-        while True:
-            if self.stop_event != None and self.stop_event.is_set():
-                return
-            try:
-                self.plasma_client = plasma.connect("/tmp/plasma")
-                return
-            except:
-                print(f"TrackedObjectProcessor: unable to connect plasma client")
-                time.sleep(10)
+    def create(self, name, size) -> AnyStr:
+        shm = shared_memory.SharedMemory(name=name, create=True, size=size)
+        self.shm_store[name] = shm
+        return shm.buf

-    def get(self, name, timeout_ms=0):
-        object_id = plasma.ObjectID(hashlib.sha1(str.encode(name)).digest())
-        while True:
-            if self.stop_event != None and self.stop_event.is_set():
-                return
-            try:
-                frame = self.plasma_client.get(object_id, timeout_ms=timeout_ms)
-                if frame is plasma.ObjectNotAvailable:
-                    return None
-                return frame
-            except:
-                self.connect()
-                time.sleep(1)
+    def get(self, name, shape):
+        if name in self.shm_store:
+            shm = self.shm_store[name]
+        else:
+            shm = shared_memory.SharedMemory(name=name)
+            self.shm_store[name] = shm
+        return np.ndarray(shape, dtype=np.uint8, buffer=shm.buf)

-    def put(self, name, frame):
-        object_id = plasma.ObjectID(hashlib.sha1(str.encode(name)).digest())
-        while True:
-            if self.stop_event != None and self.stop_event.is_set():
-                return
-            try:
-                self.plasma_client.put(frame, object_id)
-                return
-            except Exception as e:
-                print(f"Failed to put in plasma: {e}")
-                self.connect()
-                time.sleep(1)
+    def close(self, name):
+        if name in self.shm_store:
+            self.shm_store[name].close()
+            del self.shm_store[name]

    def delete(self, name):
-        object_id = plasma.ObjectID(hashlib.sha1(str.encode(name)).digest())
-        while True:
-            if self.stop_event != None and self.stop_event.is_set():
-                return
-            try:
-                self.plasma_client.delete([object_id])
-                return
-            except:
-                self.connect()
-                time.sleep(1)
+        if name in self.shm_store:
+            self.shm_store[name].close()
+            self.shm_store[name].unlink()
+            del self.shm_store[name]
--- a/frigate/video.py
+++ b/frigate/video.py
@@ -5,7 +5,6 @@ import cv2
 import queue
 import threading
 import ctypes
-import pyarrow.plasma as plasma
 import multiprocessing as mp
 import subprocess as sp
 import numpy as np
@@ -15,7 +14,7 @@ import json
 import base64
 from typing import Dict, List
 from collections import defaultdict
-from frigate.util import draw_box_with_label, area, calculate_region, clipped, intersection_over_union, intersection, EventsPerSecond, listen, FrameManager, PlasmaFrameManager
+from frigate.util import draw_box_with_label, yuv_region_2_rgb, area, calculate_region, clipped, intersection_over_union, intersection, EventsPerSecond, listen, FrameManager, SharedMemoryFrameManager
 from frigate.objects import ObjectTracker
 from frigate.edgetpu import RemoteObjectDetector
 from frigate.motion import MotionDetector
@@ -89,7 +88,7 @@ def filtered(obj, objects_to_track, object_filters, mask=None):
    return False

 def create_tensor_input(frame, region):
-    cropped_frame = frame[region[1]:region[3], region[0]:region[2]]
+    cropped_frame = yuv_region_2_rgb(frame, region)

    # Resize to 300x300 if needed
    if cropped_frame.shape != (300, 300, 3):
@@ -117,54 +116,103 @@ def start_or_restart_ffmpeg(ffmpeg_cmd, frame_size, ffmpeg_process=None):
    return process

 def capture_frames(ffmpeg_process, camera_name, frame_shape, frame_manager: FrameManager, 
-    frame_queue, take_frame: int, fps:EventsPerSecond, skipped_fps: EventsPerSecond, 
-    stop_event: mp.Event, detection_frame: mp.Value, current_frame: mp.Value):
+    frame_queue, take_frame: int, fps:mp.Value, skipped_fps: mp.Value, 
+    stop_event: mp.Event, current_frame: mp.Value):

    frame_num = 0
-    last_frame = 0
-    frame_size = frame_shape[0] * frame_shape[1] * frame_shape[2]
-    skipped_fps.start()
+    frame_size = frame_shape[0] * frame_shape[1] * 3 // 2
+    frame_rate = EventsPerSecond()
+    skipped_eps = EventsPerSecond()
+    skipped_eps.start()
    while True:
+        fps.value = frame_rate.eps()
+        skipped_fps = skipped_eps.eps()
        if stop_event.is_set():
            print(f"{camera_name}: stop event set. exiting capture thread...")
            break

-        frame_bytes = ffmpeg_process.stdout.read(frame_size)
        current_frame.value = datetime.datetime.now().timestamp()
+        frame_name = f"{camera_name}{current_frame.value}"
+        frame_buffer = frame_manager.create(frame_name, frame_size)
+        try:
+          frame_buffer[:] = ffmpeg_process.stdout.read(frame_size)
+        except:
+          print(f"{camera_name}: ffmpeg sent a broken frame. something is wrong.")

-        if len(frame_bytes) < frame_size:
-            print(f"{camera_name}: ffmpeg sent a broken frame. something is wrong.")
+          if ffmpeg_process.poll() != None:
+              print(f"{camera_name}: ffmpeg process is not running. exiting capture thread...")
+              frame_manager.delete(frame_name)
+              break
+          
+          continue

-            if ffmpeg_process.poll() != None:
-                print(f"{camera_name}: ffmpeg process is not running. exiting capture thread...")
-                break
-            else:
-                continue
-
-        fps.update()
+        frame_rate.update()

        frame_num += 1
        if (frame_num % take_frame) != 0:
-            skipped_fps.update()
+            skipped_eps.update()
+            frame_manager.delete(frame_name)
            continue

-        # if the detection process is more than 1 second behind, skip this frame
-        if detection_frame.value > 0.0 and (last_frame - detection_frame.value) > 1:
-            skipped_fps.update()
+        # if the queue is full, skip this frame
+        if frame_queue.full():
+            skipped_eps.update()
+            frame_manager.delete(frame_name)
            continue

-        # put the frame in the frame manager
-        frame_manager.put(f"{camera_name}{current_frame.value}",
-                np
-                    .frombuffer(frame_bytes, np.uint8)
-                    .reshape(frame_shape)
-            )
+        # close the frame
+        frame_manager.close(frame_name)
+
        # add to the queue
        frame_queue.put(current_frame.value)
-        last_frame = current_frame.value
+
+class CameraWatchdog(threading.Thread):
+    def __init__(self, name, config, frame_queue, camera_fps, stop_event):
+        threading.Thread.__init__(self)
+        self.name = name
+        self.config = config
+        self.capture_thread = None
+        self.ffmpeg_process = None
+        self.stop_event = stop_event
+        self.camera_fps = camera_fps
+        self.frame_queue = frame_queue
+        self.frame_shape = self.config['frame_shape']
+        self.frame_size = self.frame_shape[0] * self.frame_shape[1] * 3 // 2
+
+    def run(self):
+        self.start_ffmpeg()
+        time.sleep(10)
+        while True:
+            if self.stop_event.is_set():
+                print(f"Exiting watchdog...")
+                break
+
+            now = datetime.datetime.now().timestamp()
+
+            if not self.capture_thread.is_alive():
+                self.start_ffmpeg()
+            elif now - self.capture_thread.current_frame.value > 5:
+                print(f"No frames received from {self.name} in 5 seconds. Exiting ffmpeg...")
+                self.ffmpeg_process.terminate()
+                try:
+                    print("Waiting for ffmpeg to exit gracefully...")
+                    self.ffmpeg_process.communicate(timeout=30)
+                except sp.TimeoutExpired:
+                    print("FFmpeg didnt exit. Force killing...")
+                    self.ffmpeg_process.kill()
+                    self.ffmpeg_process.communicate()
+            
+            # wait a bit before checking again
+            time.sleep(10)
+    
+    def start_ffmpeg(self):
+      self.ffmpeg_process = start_or_restart_ffmpeg(self.config['ffmpeg_cmd'], self.frame_size)
+      self.capture_thread = CameraCapture(self.name, self.ffmpeg_process, self.frame_shape, self.frame_queue, 
+          self.config['take_frame'], self.camera_fps, self.stop_event)
+      self.capture_thread.start()

 class CameraCapture(threading.Thread):
-    def __init__(self, name, ffmpeg_process, frame_shape, frame_queue, take_frame, fps, detection_frame, stop_event):
+    def __init__(self, name, ffmpeg_process, frame_shape, frame_queue, take_frame, fps, stop_event):
        threading.Thread.__init__(self)
        self.name = name
        self.frame_shape = frame_shape
@@ -173,23 +221,29 @@ class CameraCapture(threading.Thread):
        self.take_frame = take_frame
        self.fps = fps
        self.skipped_fps = EventsPerSecond()
-        self.plasma_client = PlasmaFrameManager(stop_event)
+        self.frame_manager = SharedMemoryFrameManager()
        self.ffmpeg_process = ffmpeg_process
        self.current_frame = mp.Value('d', 0.0)
        self.last_frame = 0
-        self.detection_frame = detection_frame
        self.stop_event = stop_event

    def run(self):
        self.skipped_fps.start()
-        capture_frames(self.ffmpeg_process, self.name, self.frame_shape, self.plasma_client, self.frame_queue, self.take_frame,
-            self.fps, self.skipped_fps, self.stop_event, self.detection_frame, self.current_frame)
+        capture_frames(self.ffmpeg_process, self.name, self.frame_shape, self.frame_manager, self.frame_queue, self.take_frame,
+            self.fps, self.skipped_fps, self.stop_event, self.current_frame)

-def track_camera(name, config, frame_queue, frame_shape, detection_queue, detected_objects_queue, fps, detection_fps, read_start, detection_frame, stop_event):
-    print(f"Starting process for {name}: {os.getpid()}")
+def capture_camera(name, config, process_info, stop_event):
+    frame_queue = process_info['frame_queue']
+    camera_watchdog = CameraWatchdog(name, config, frame_queue, process_info['camera_fps'], stop_event)
+    camera_watchdog.start()
+    camera_watchdog.join()
+
+def track_camera(name, config, detection_queue, result_connection, detected_objects_queue, process_info, stop_event):
    listen()

-    detection_frame.value = 0.0
+    frame_queue = process_info['frame_queue']
+
+    frame_shape = config['frame_shape']

    # Merge the tracked object config with the global config
    camera_objects_config = config.get('objects', {})
@@ -218,14 +272,14 @@ def track_camera(name, config, frame_queue, frame_shape, detection_queue, detect
        mask[:] = 255

    motion_detector = MotionDetector(frame_shape, mask, resize_factor=6)
-    object_detector = RemoteObjectDetector(name, '/labelmap.txt', detection_queue)
+    object_detector = RemoteObjectDetector(name, '/labelmap.txt', detection_queue, result_connection)

    object_tracker = ObjectTracker(10)

-    plasma_client = PlasmaFrameManager()
+    frame_manager = SharedMemoryFrameManager()

-    process_frames(name, frame_queue, frame_shape, plasma_client, motion_detector, object_detector,
-        object_tracker, detected_objects_queue, fps, detection_fps, detection_frame, objects_to_track, object_filters, mask, stop_event)
+    process_frames(name, frame_queue, frame_shape, frame_manager, motion_detector, object_detector,
+        object_tracker, detected_objects_queue, process_info, objects_to_track, object_filters, mask, stop_event)

    print(f"{name}: exiting subprocess")

@@ -261,10 +315,14 @@ def detect(object_detector, frame, region, objects_to_track, object_filters, mas
 def process_frames(camera_name: str, frame_queue: mp.Queue, frame_shape, 
    frame_manager: FrameManager, motion_detector: MotionDetector, 
    object_detector: RemoteObjectDetector, object_tracker: ObjectTracker,
-    detected_objects_queue: mp.Queue, fps: mp.Value, detection_fps: mp.Value, current_frame_time: mp.Value,
+    detected_objects_queue: mp.Queue, process_info: Dict,
    objects_to_track: List[str], object_filters: Dict, mask, stop_event: mp.Event,
    exit_on_empty: bool = False):
    
+    fps = process_info['process_fps']
+    detection_fps = process_info['detection_fps']
+    current_frame_time = process_info['detection_frame']
+
    fps_tracker = EventsPerSecond()
    fps_tracker.start()

@@ -278,17 +336,13 @@ def process_frames(camera_name: str, frame_queue: mp.Queue, frame_shape,
        except queue.Empty:
            continue

-        
        current_frame_time.value = frame_time

-        frame = frame_manager.get(f"{camera_name}{frame_time}")
+        frame = frame_manager.get(f"{camera_name}{frame_time}", (frame_shape[0]*3//2, frame_shape[1]))

        if frame is None:
            print(f"{camera_name}: frame {frame_time} is not in memory store.")
            continue
-        
-        fps_tracker.update()
-        fps.value = fps_tracker.eps()

        # look for motion
        motion_boxes = motion_detector.detect(frame)
@@ -308,7 +362,7 @@ def process_frames(camera_name: str, frame_queue: mp.Queue, frame_shape,
        # re-compute regions
        regions = [calculate_region(frame_shape, a[0], a[1], a[2], a[3], 1.0)
            for a in combined_regions]
-        
+
        # resize regions and detect
        detections = []
        for region in regions:
@@ -360,7 +414,13 @@ def process_frames(camera_name: str, frame_queue: mp.Queue, frame_shape,
        # now that we have refined our detections, we need to track objects
        object_tracker.match_and_update(frame_time, detections)

-        # add to the queue
-        detected_objects_queue.put((camera_name, frame_time, object_tracker.tracked_objects))
-
-        detection_fps.value = object_detector.fps.eps()
+        # add to the queue if not full
+        if(detected_objects_queue.full()):
+          frame_manager.delete(f"{camera_name}{frame_time}")
+          continue
+        else:
+          fps_tracker.update()
+          fps.value = fps_tracker.eps()
+          detected_objects_queue.put((camera_name, frame_time, object_tracker.tracked_objects))
+          detection_fps.value = object_detector.fps.eps()
+          frame_manager.close(f"{camera_name}{frame_time}")
--- a/process_clip.py
+++ b/process_clip.py
@@ -4,7 +4,7 @@ import os
 import datetime
 from unittest import TestCase, main
 from frigate.video import process_frames, start_or_restart_ffmpeg, capture_frames, get_frame_shape
-from frigate.util import DictFrameManager, EventsPerSecond, draw_box_with_label
+from frigate.util import DictFrameManager, SharedMemoryFrameManager, EventsPerSecond, draw_box_with_label
 from frigate.motion import MotionDetector
 from frigate.edgetpu import LocalObjectDetector
 from frigate.objects import ObjectTracker
@@ -19,6 +19,7 @@ class ProcessClip():
        self.frame_shape = frame_shape
        self.camera_name = 'camera'
        self.frame_manager = DictFrameManager()
+        # self.frame_manager = SharedMemoryFrameManager()
        self.frame_queue = mp.Queue()
        self.detected_objects_queue = mp.Queue()
        self.camera_state = CameraState(self.camera_name, config, self.frame_manager)
@@ -72,13 +73,15 @@ class ProcessClip():
            for obj in self.camera_state.tracked_objects.values():
                print(f"{frame_time}: {obj['id']} - {obj['computed_score']} - {obj['score_history']}")
        
+        self.frame_manager.delete(self.camera_state.previous_frame_id)
+        
        return {
            'object_detected': obj_detected,
            'top_score': top_computed_score
        }
    
    def save_debug_frame(self, debug_path, frame_time, tracked_objects):
-        current_frame = self.frame_manager.get(f"{self.camera_name}{frame_time}")
+        current_frame = self.frame_manager.get(f"{self.camera_name}{frame_time}", self.frame_shape)
        # draw the bounding boxes on the frame
        for obj in tracked_objects:
            thickness = 2
@@ -132,6 +135,7 @@ def process(path, label, threshold, debug_path):
    results = []
    for c in clips:
        frame_shape = get_frame_shape(c)
+        config['frame_shape'] = frame_shape
        process_clip = ProcessClip(c, frame_shape, config)
        process_clip.load_frames()
        process_clip.process_frames(objects_to_track=config['objects']['track'])
Author	SHA1	Message	Date
Blake Blackshear	b6f7940b10	hwaccel docs	2020-10-25 14:30:36 -05:00
Blake Blackshear	75312602aa	add support for iHD driver	2020-10-25 14:30:36 -05:00
Blake Blackshear	75977128f0	ensure dummy frame is in yuv shape	2020-10-25 14:30:36 -05:00
Blake Blackshear	eafde6c677	capture ffmpeg in a dedicated process	2020-10-25 14:30:36 -05:00
Blake Blackshear	da0598baef	disable flask warning	2020-10-25 14:30:36 -05:00
Blake Blackshear	35ba5e2f7c	improve frame memory management	2020-10-25 14:30:36 -05:00
Blake Blackshear	49258d6dbe	tweaks for recent issues	2020-10-24 08:52:40 -05:00
Blake Blackshear	5a081e4f00	docs rewrite	2020-10-24 08:23:16 -05:00
Blake Blackshear	4feae472e9	reformatting and fixing typos	2020-10-23 06:56:06 -05:00
tubalainen	4e83239258	Updated information on poly mask	2020-10-23 06:56:06 -05:00
tubalainen	c4cccf44a5	poly example image	2020-10-23 06:38:41 -05:00
jacobgibbs	64e7cbcc62	Update README.md Update attributes name to pull through the FPS	2020-10-19 15:04:34 -05:00
Blake Blackshear	dd86e4f317	fix clips path and check for symlinks	2020-10-19 07:01:31 -05:00
Blake Blackshear	4db285a875	remove reference to stable	2020-10-18 14:12:25 -05:00
Blake Blackshear	939d1ba091	use global and ensure dirs exist	2020-10-18 13:47:13 -05:00
Blake Blackshear	0fe8d486d9	make cache/clips dirs configurable	2020-10-18 13:47:13 -05:00
Blake Blackshear	a3cb02af5c	sync arch names with hassio	2020-10-18 13:47:13 -05:00
Blake Blackshear	45a6b8452c	allow config file to be specified by env var and allow json	2020-10-18 13:47:13 -05:00
Blake Blackshear	9d594cc640	allow setting config file location via env var	2020-10-18 13:47:13 -05:00
Blake Blackshear	59e41ae1ac	update sample config	2020-10-18 13:47:13 -05:00
Blake Blackshear	c6ed16465b	move the timestamp to bottom	2020-10-18 13:47:13 -05:00
Blake Blackshear	8f14b36f5a	tweak size	2020-10-18 13:47:13 -05:00
Blake Blackshear	b6c2491e3b	use the actual original shape	2020-10-18 13:47:13 -05:00
Blake Blackshear	8e31d04d90	scale font of timestamp dynamically	2020-10-18 13:47:13 -05:00
Blake Blackshear	bf93fbb357	add ability to draw bounding boxes/timestamps on snapshots	2020-10-18 13:47:13 -05:00
Blake Blackshear	c064b244db	handle empty best frames	2020-10-18 13:47:13 -05:00
Blake Blackshear	0280610e96	fix detector cleanup	2020-10-18 13:47:13 -05:00
Blake Blackshear	4363623c45	reduce zone filter bouncing	2020-10-18 13:47:13 -05:00
Blake Blackshear	c960914ec3	prevent the camera process from hanging	2020-10-18 13:47:13 -05:00
Blake Blackshear	9ecc80b443	syntax error	2020-10-18 13:47:13 -05:00
Blake Blackshear	3e146de0a2	update docs	2020-10-18 13:47:13 -05:00
Blake Blackshear	bee54c39dc	update default detectors	2020-10-18 13:47:13 -05:00
Blake Blackshear	623d138d60	use dictionary for detectors for sensors	2020-10-18 13:47:13 -05:00
Blake Blackshear	76befc1249	only draw during debug	2020-10-18 13:47:13 -05:00
Dejan Zelic	51251b9fb0	Added Healthcheck to Docker Compose Frigate provides an HTTP server that can be used to detect if frigate is running or not. Using the docker-compose "healthcheck" feature we can set automations to restart the service if it stops working.	2020-10-18 13:47:13 -05:00
Radegast	8c45076bb6	Fix error in the docker run command I have very little experience with Docker, but it seems the command in the README has two mistakes in it: - unknown shorthand flag: 'n' in -name - docker: Error response from daemon: Invalid container name (blakeblackshear/frigate:stable), only [a-zA-Z0-9][a-zA-Z0-9_.-] are allowed. I am running Docker version 19.03.13-ce, build 4484c46d9d on Arch linux.	2020-10-18 13:47:13 -05:00
Blake Blackshear	7d683ef399	cleanup frame queue	2020-10-18 13:47:13 -05:00
Blake Blackshear	e4da3822b1	cleanup detection shms	2020-10-18 13:47:13 -05:00
Blake Blackshear	12c4cd77c5	only convert pix_fmt when necessary	2020-10-18 13:47:13 -05:00
Blake Blackshear	a611cbb942	use yuv420p pixel format for motion	2020-10-18 13:47:13 -05:00
Blake Blackshear	f946813ccb	support multiple coral devices (fixes #100 )	2020-10-18 13:47:13 -05:00
Blake Blackshear	49fca1b839	print stacktraceon segfaults	2020-10-18 13:47:13 -05:00
Blake Blackshear	54cb4a2180	prevent frame from being deleted while in use	2020-10-18 13:47:13 -05:00
Blake Blackshear	9954e3b11e	build ffmpeg in separate container	2020-10-18 13:47:13 -05:00
Blake Blackshear	82692b0ddc	arm64 ffmpeg cleanup	2020-10-18 13:47:13 -05:00
Blake Blackshear	9d4fdec12f	arm64 ffmpeg build	2020-10-18 13:47:13 -05:00
Blake Blackshear	ed72c995ef	ffmpeg 4.3.1 build for amd64	2020-10-18 13:47:13 -05:00
Blake Blackshear	66c77d1157	base image build cleanup	2020-10-18 13:47:13 -05:00
Blake Blackshear	40c322ad47	arm64 support	2020-10-18 13:47:13 -05:00
Blake Blackshear	83f1e0d713	add rpi dockerfile	2020-10-18 13:47:13 -05:00
Blake Blackshear	2d89044bd3	update dockerfiles for amd64	2020-10-18 13:47:13 -05:00
Blake Blackshear	dc4d24c2b9	Base dockerfile for building wheels	2020-10-18 13:47:13 -05:00
Blake Blackshear	d5fb20c524	refactor dockerfile	2020-10-18 13:47:13 -05:00
Blake Blackshear	7e92e8bfe8	fix shared memory store usage for events	2020-10-18 13:47:13 -05:00
Blake Blackshear	efdcfcef97	cleanup	2020-10-18 13:47:13 -05:00
Blake Blackshear	574ee2a46f	update detection handoff to use shared memory	2020-10-18 13:47:13 -05:00
Blake Blackshear	ec4d048905	upgrade to python3.8 and switch from plasma store to shared_memory	2020-10-18 13:47:13 -05:00