free_network_ptr instead of free_network

Examples

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,26 @@
+---
+name: Bug report
+about: Create a report to help us improve
+title: "[BUG]"
+labels: bug, help wanted
+assignees: LdDl
+
+---
+
+**Describe the bug**
+A clear and concise description of what the bug is.
+
+**To Reproduce**
+Steps to reproduce the behavior
+
+**Expected behavior**
+A clear and concise description of what you expected to happen.
+
+**Expected behavior**
+A clear and concise description of what you expected to happen.
+
+**Describe the solution you'd like and provide pseudocode examples if you can**
+A clear and concise description of what you want to happen.
+
+**Additional context**
+Add any other context about the problem here.

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,20 @@
+---
+name: Feature request
+about: Suggest an idea for this project
+title: "[FEATURE REQUEST]"
+labels: enhancement
+assignees: LdDl
+
+---
+
+**Is your feature request related to a problem? Please describe.**
+A clear and concise description of what the problem is.
+
+**Describe the solution you'd like and provide pseudocode examples if you can**
+A clear and concise description of what you want to happen.
+
+**Describe alternatives you've considered and provide pseudocode examples if you can**
+A clear and concise description of any alternative solutions or features you've considered.
+
+**Additional context**
+Add any other context or screenshots about the feature request here.

.gitignore

@@ -1,5 +1,4 @@
 example/main
-example/sample.jpg
 example/coco.names
 example/yolov3.cfg
 example/yolov3.weights

Makefile

@@ -0,0 +1,55 @@
+.ONESHELL:
+.PHONY: download build clean
+
+# Latest battletested AlexeyAB version of Darknet commit
+LATEST_COMMIT?=f056fc3b6a11528fa0522a468eca1e909b7004b7
+
+# Temporary folder for building Darknet
+TMP_DIR?=/tmp/
+
+# Download AlexeyAB version of Darknet
+download:
+	rm -rf $(TMP_DIR)install_darknet
+	mkdir $(TMP_DIR)install_darknet
+	git clone https://github.com/AlexeyAB/darknet.git $(TMP_DIR)install_darknet
+	cd $(TMP_DIR)install_darknet
+	git checkout $(LATEST_COMMIT)
+	cd -
+
+# Build AlexeyAB version of Darknet for usage with CPU only.
+build:
+	cd $(TMP_DIR)install_darknet
+	sed -i -e 's/GPU=1/GPU=0/g' Makefile
+	sed -i -e 's/CUDNN=1/CUDNN=0/g' Makefile
+	sed -i -e 's/LIBSO=0/LIBSO=1/g' Makefile
+	$(MAKE) -j $(shell nproc --all)
+	$(MAKE) preinstall
+	cd -
+
+# Build AlexeyAB version of Darknet for usage with both CPU and GPU (CUDA by NVIDIA).
+build_gpu:
+	cd $(TMP_DIR)install_darknet
+	sed -i -e 's/GPU=0/GPU=1/g' Makefile
+	sed -i -e 's/CUDNN=0/CUDNN=1/g' Makefile
+	sed -i -e 's/LIBSO=0/LIBSO=1/g' Makefile
+	$(MAKE) -j $(shell nproc --all)
+	$(MAKE) preinstall
+	cd -
+
+# Install system wide.
+sudo_install:
+	cd $(TMP_DIR)install_darknet
+	sudo cp libdarknet.so /usr/lib/libdarknet.so
+	sudo cp include/darknet.h /usr/include/darknet.h
+	sudo ldconfig
+	cd -
+
+# Cleanup temporary files for building process
+clean:
+	rm -rf $(TMP_DIR)install_darknet
+
+# Do every step for CPU-based only build.
+install: download build sudo_install clean
+
+# Do every step for both CPU and GPU-based build.
+install_gpu: download build_gpu sudo_install clean

README.md

@@ -3,7 +3,6 @@
 [![Go Report Card](https://goreportcard.com/badge/github.com/LdDl/go-darknet)](https://goreportcard.com/report/github.com/LdDl/go-darknet)
 [![GitHub tag](https://img.shields.io/github/tag/LdDl/go-darknet.svg)](https://github.com/LdDl/go-darknet/releases)
 
-
 # go-darknet: Go bindings for Darknet (Yolo V4, Yolo V3)
 ### go-darknet is a Go package, which uses Cgo to enable Go applications to use YOLO V4/V3 in [Darknet].
 
@@ -23,29 +22,20 @@
 
 ## Requirements
 
-For proper codebase please use fork of [darknet](https://github.com/AlexeyAB/darknet). Latest commit I've tested [here](https://github.com/AlexeyAB/darknet/commit/9dc897d2c77d5ef43a6b237b717437375765b527)
+You need to install fork of [darknet](https://github.com/AlexeyAB/darknet). Latest commit I've tested is [here](https://github.com/AlexeyAB/darknet/commit/d65909fbea471d06e52a2e4a41132380dc2edaa6)
 
-In order to use go-darknet, `libdarknet.so` should be available in one of
-the following locations:
+Use provided [Makefile](Makefile).
 
-* /usr/lib
-* /usr/local/lib
-
-Also, [darknet.h] should be available in one of the following locations:
-
-* /usr/include
-* /usr/local/include
-
-To achieve it, after Darknet compilation (via make) execute following command:
+* For CPU-based instalattion:
     ```shell
-# Copy *.so to /usr/lib + /usr/include (or /usr/local/lib + /usr/local/include)
-sudo cp libdarknet.so /usr/lib/libdarknet.so && sudo cp include/darknet.h /usr/include/darknet.h
-# sudo cp libdarknet.so /usr/local/lib/libdarknet.so && sudo cp include/darknet.h /usr/local/include/darknet.h
+    make install
     ```
-Note: do not forget to set LIBSO=1 in Makefile before executing 'make':
-```Makefile
-LIBSO=1
+* For both CPU and GPU-based instalattion:
+    ```shell
+    make install_gpu
     ```
+    Note: If you want to have GPU-acceleration before running command above install [CUDA](https://developer.nvidia.com/cuda-downloads) and [cuDNN](https://developer.nvidia.com/cudnn) (Latest CUDA version I've tested is [10.2](https://developer.nvidia.com/cuda-10.2-download-archive) and cuDNN is [7.6.5](https://developer.nvidia.com/rdp/cudnn-archive#a-collapse765-102))
+
 ## Installation
 
 ```shell
@@ -58,19 +48,19 @@ Example Go program is provided in the [example] directory. Please refer to the c
 
 Building and running program:
 
-Navigate to [example] folder
+* Navigate to [example] folder
     ```shell
-cd $GOPATH/github.com/LdDl/go-darknet/example
+    cd $GOPATH/github.com/LdDl/go-darknet/example/base_example
     ```
 
-Download dataset (sample of image, coco.names, yolov4.cfg (or v3), yolov4.weights(or v3)).
+* Download dataset (sample of image, coco.names, yolov4.cfg (or v3), yolov4.weights(or v3)).
     ```shell
     #for yolo v4
     ./download_data.sh
     #for yolo v3
     ./download_data_v3.sh
     ```
-Note: you don't need *coco.data* file anymore, because sh-script above does insert *coco.names* into 'names' field in *yolov4.cfg* file (so AlexeyAB's fork can deal with it properly)
+* Note: you don't need *coco.data* file anymore, because sh-script above does insert *coco.names* into 'names' field in *yolov4.cfg* file (so AlexeyAB's fork can deal with it properly)
     So last rows in yolov4.cfg file will look like:
     ```bash
     ......
@@ -82,7 +72,8 @@ beta_nms=0.6
 
     names = coco.names # this is path to coco.names file
     ```
-Also do not forget change batch and subdivisions sizes from:
+
+* Also do not forget change batch and subdivisions sizes from:
     ```shell
     batch=64
     subdivisions=8
@@ -95,7 +86,7 @@ subdivisions=1
     It will reduce amount of VRAM used for detector test.
 
 
-Build and run program
+* Build and run program
     Yolo V4:
     ```shell
     go build main.go && ./main --configFile=yolov4.cfg --weightsFile=yolov4.weights --imageFile=sample.jpg
@@ -154,5 +145,5 @@ go-darknet follows [Darknet]'s [license].
 [darknet.h]: https://github.com/AlexeyAB/darknet/blob/master/include/darknet.h
 [include/darknet.h]: https://github.com/AlexeyAB/darknet/blob/master/include/darknet.h
 [Makefile]: https://github.com/alexeyab/darknet/blob/master/Makefile
-[example]: /example
+[example]: /example/base_example
 [GoDoc]: https://godoc.org/github.com/LdDl/go-darknet

example/base_example/README.md

@@ -8,12 +8,12 @@ This is an example Go application which uses go-darknet.
 Navigate to example folder:
 
 ```shell
-cd $GOPATH/github.com/LdDl/go-darknet/example
+cd $GOPATH/github.com/LdDl/go-darknet/example/base_example
 ```
 
 Download dataset (sample of image, coco.names, yolov3.cfg, yolov3.weights).
 ```shell
-./download_data.sh
+./download_data_v3.sh
 ```
 Note: you don't need *coco.data* file anymore, because script below does insert *coco.names* into 'names' filed in *yolov3.cfg* file (so AlexeyAB's fork can deal with it properly)
 So last rows in yolov3.cfg file will look like:

example/base_example/main.go

@@ -9,7 +9,6 @@ import (
 	"log"
 	"math"
 	"os"
-
 	darknet "github.com/LdDl/go-darknet"
 	"github.com/disintegration/imaging"
 )
@@ -61,13 +60,13 @@ func main() {
 	if err != nil {
 		panic(err.Error())
 	}
-	defer imgDarknet.Close()
 	
 	dr, err := n.Detect(imgDarknet)
 	if err != nil {
 		printError(err)
 		return
 	}
+	imgDarknet.Close()
 
 	log.Println("Network-only time taken:", dr.NetworkOnlyTimeTaken)
 	log.Println("Overall time taken:", dr.OverallTimeTaken, len(dr.Detections))
@@ -92,6 +91,8 @@ func main() {
 			// }
 		}
 	}
+
+	n.Close()
 }
 
 func imageToBytes(img image.Image) ([]byte, error) {

example/rest_example/README.md

@@ -0,0 +1,221 @@
+# Example Go application using go-darknet and REST
+
+This is an example Go server application (in terms of REST) which uses go-darknet.
+
+## Run
+
+Navigate to example folder:
+
+```shell
+cd $GOPATH/github.com/LdDl/go-darknet/example/rest_example
+```
+
+Download dataset (sample of image, coco.names, yolov3.cfg, yolov3.weights).
+```shell
+./download_data_v3.sh
+```
+Note: you don't need *coco.data* file anymore, because script below does insert *coco.names* into 'names' filed in *yolov3.cfg* file (so AlexeyAB's fork can deal with it properly)
+So last rows in yolov3.cfg file will look like:
+```bash
+......
+[yolo]
+mask = 0,1,2
+anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
+classes=80
+num=9
+jitter=.3
+ignore_thresh = .7
+truth_thresh = 1
+random=1
+names = coco.names # this is path to coco.names file
+```
+
+Build and run program
+```
+go build main.go && ./main --configFile=yolov3.cfg --weightsFile=yolov3.weights --port 8090
+```
+
+After server started check if REST-requests works. We provide cURL-based example
+```shell
+curl -F 'image=@sample.jpg' 'http://localhost:8090/detect_objects'
+```
+
+Servers response should be something like this:
+```json
+{
+    "net_time": "43.269289ms",
+    "overall_time": "43.551604ms",
+    "num_detections": 44,
+    "detections": [
+        {
+            "class_id": 7,
+            "class_name": "truck",
+            "probability": 49.51231,
+            "start_point": {
+                "x": 0,
+                "y": 136
+            },
+            "end_point": {
+                "x": 85,
+                "y": 311
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 36.36933,
+            "start_point": {
+                "x": 95,
+                "y": 152
+            },
+            "end_point": {
+                "x": 186,
+                "y": 283
+            }
+        },
+        {
+            "class_id": 7,
+            "class_name": "truck",
+            "probability": 48.417683,
+            "start_point": {
+                "x": 95,
+                "y": 152
+            },
+            "end_point": {
+                "x": 186,
+                "y": 283
+            }
+        },
+        {
+            "class_id": 7,
+            "class_name": "truck",
+            "probability": 45.652023,
+            "start_point": {
+                "x": 694,
+                "y": 178
+            },
+            "end_point": {
+                "x": 798,
+                "y": 310
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 76.8402,
+            "start_point": {
+                "x": 1,
+                "y": 145
+            },
+            "end_point": {
+                "x": 84,
+                "y": 324
+            }
+        },
+        {
+            "class_id": 7,
+            "class_name": "truck",
+            "probability": 25.592052,
+            "start_point": {
+                "x": 107,
+                "y": 89
+            },
+            "end_point": {
+                "x": 215,
+                "y": 263
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 99.87823,
+            "start_point": {
+                "x": 511,
+                "y": 185
+            },
+            "end_point": {
+                "x": 748,
+                "y": 328
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 99.819336,
+            "start_point": {
+                "x": 261,
+                "y": 189
+            },
+            "end_point": {
+                "x": 427,
+                "y": 322
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 99.64055,
+            "start_point": {
+                "x": 426,
+                "y": 197
+            },
+            "end_point": {
+                "x": 539,
+                "y": 311
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 74.56263,
+            "start_point": {
+                "x": 692,
+                "y": 186
+            },
+            "end_point": {
+                "x": 796,
+                "y": 316
+            }
+        },
+        {
+            "class_id": 2,
+            "class_name": "car",
+            "probability": 72.79756,
+            "start_point": {
+                "x": 388,
+                "y": 206
+            },
+            "end_point": {
+                "x": 437,
+                "y": 276
+            }
+        },
+        {
+            "class_id": 1,
+            "class_name": "bicycle",
+            "probability": 72.27595,
+            "start_point": {
+                "x": 178,
+                "y": 270
+            },
+            "end_point": {
+                "x": 268,
+                "y": 406
+            }
+        },
+        {
+            "class_id": 0,
+            "class_name": "person",
+            "probability": 97.30075,
+            "start_point": {
+                "x": 143,
+                "y": 135
+            },
+            "end_point": {
+                "x": 268,
+                "y": 343
+            }
+        }
+    ]
+}
+```

example/rest_example/main.go

@@ -0,0 +1,140 @@
+package main
+
+import (
+	"bytes"
+	"encoding/json"
+	"flag"
+	"fmt"
+	"image"
+	_ "image/jpeg"
+	"io/ioutil"
+	"log"
+	"net/http"
+
+	"github.com/LdDl/go-darknet"
+)
+
+var configFile = flag.String("configFile", "",
+	"Path to network layer configuration file. Example: cfg/yolov3.cfg")
+var weightsFile = flag.String("weightsFile", "",
+	"Path to weights file. Example: yolov3.weights")
+var serverPort = flag.Int("port", 8090,
+	"Listening port")
+
+func main() {
+	flag.Parse()
+
+	if *configFile == "" || *weightsFile == "" {
+		flag.Usage()
+		return
+	}
+
+	n := darknet.YOLONetwork{
+		GPUDeviceIndex:           0,
+		NetworkConfigurationFile: *configFile,
+		WeightsFile:              *weightsFile,
+		Threshold:                .25,
+	}
+	if err := n.Init(); err != nil {
+		log.Println(err)
+		return
+	}
+	defer n.Close()
+
+	http.HandleFunc("/detect_objects", detectObjects(&n))
+	http.ListenAndServe(fmt.Sprintf(":%d", *serverPort), nil)
+}
+
+// DarknetResp Response
+type DarknetResp struct {
+	NetTime     string              `json:"net_time"`
+	OverallTime string              `json:"overall_time"`
+	Detections  []*DarknetDetection `json:"detections"`
+}
+
+// DarknetDetection Information about single detection
+type DarknetDetection struct {
+	ClassID     int           `json:"class_id"`
+	ClassName   string        `json:"class_name"`
+	Probability float32       `json:"probability"`
+	StartPoint  *DarknetPoint `json:"start_point"`
+	EndPoint    *DarknetPoint `json:"end_point"`
+}
+
+// DarknetPoint image.Image point
+type DarknetPoint struct {
+	X int `json:"x"`
+	Y int `json:"y"`
+}
+
+func detectObjects(n *darknet.YOLONetwork) func(w http.ResponseWriter, req *http.Request) {
+	return func(w http.ResponseWriter, req *http.Request) {
+		// Restrict file size up to 10mb
+		req.ParseMultipartForm(10 << 20)
+
+		file, _, err := req.FormFile("image")
+		if err != nil {
+			fmt.Fprintf(w, fmt.Sprintf("Error reading FormFile: %s", err.Error()))
+			return
+		}
+		defer file.Close()
+
+		fileBytes, err := ioutil.ReadAll(file)
+		if err != nil {
+			fmt.Fprintf(w, fmt.Sprintf("Error reading bytes: %s", err.Error()))
+			return
+		}
+
+		imgSrc, _, err := image.Decode(bytes.NewReader(fileBytes))
+		if err != nil {
+			fmt.Fprintf(w, fmt.Sprintf("Error decoding bytes to image: %s", err.Error()))
+			return
+		}
+
+		imgDarknet, err := darknet.Image2Float32(imgSrc)
+		if err != nil {
+			fmt.Fprintf(w, fmt.Sprintf("Error converting image.Image to darknet.DarknetImage: %s", err.Error()))
+			return
+		}
+		defer imgDarknet.Close()
+
+		dr, err := n.Detect(imgDarknet)
+		if err != nil {
+			fmt.Fprintf(w, fmt.Sprintf("Error detecting objects: %s", err.Error()))
+			return
+		}
+
+		resp := DarknetResp{
+			NetTime:     fmt.Sprintf("%v", dr.NetworkOnlyTimeTaken),
+			OverallTime: fmt.Sprintf("%v", dr.OverallTimeTaken),
+			Detections:  []*DarknetDetection{},
+		}
+
+		for _, d := range dr.Detections {
+			for i := range d.ClassIDs {
+				bBox := d.BoundingBox
+				resp.Detections = append(resp.Detections, &DarknetDetection{
+					ClassID:     d.ClassIDs[i],
+					ClassName:   d.ClassNames[i],
+					Probability: d.Probabilities[i],
+					StartPoint: &DarknetPoint{
+						X: bBox.StartPoint.X,
+						Y: bBox.StartPoint.Y,
+					},
+					EndPoint: &DarknetPoint{
+						X: bBox.EndPoint.X,
+						Y: bBox.EndPoint.Y,
+					},
+				})
+			}
+		}
+
+		respBytes, err := json.Marshal(resp)
+		if err != nil {
+			fmt.Fprintf(w, fmt.Sprintf("Error encoding response: %s", err.Error()))
+			return
+		}
+
+		fmt.Fprintf(w, string(respBytes))
+	}
+}

network.c

@@ -14,7 +14,8 @@ struct network_box_result perform_network_detect(network *n, image *img, int cla
         sized = resize_image(*img, n->w, n->h);
     }
     struct network_box_result result = { NULL };
-    network_predict(*n, sized.data);
+    // mleak at network_predict(), get_network_boxes() and network_predict_ptr()?
+    network_predict_ptr(n, sized.data);
     int nboxes = 0;
     result.detections = get_network_boxes(n, img->w, img->h, thresh, hier_thresh, 0, 1, &result.detections_len, letter_box);
     if (nms) {

network.go

@@ -58,7 +58,7 @@ func (n *YOLONetwork) Close() error {
 	if n.cNet == nil {
 		return errNetworkNotInit
 	}
-	C.free_network(*n.cNet)
+	C.free_network_ptr(n.cNet)
 	n.cNet = nil
 	return nil
 }