mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-05 16:48:03 +08:00
Add docs for external models (#95)
* first commit for yolov7 * pybind for yolov7 * CPP README.md * CPP README.md * modified yolov7.cc * README.md * python file modify * delete license in fastdeploy/ * repush the conflict part * README.md modified * README.md modified * file path modified * file path modified * file path modified * file path modified * file path modified * README modified * README modified * move some helpers to private * add examples for yolov7 * api.md modified * api.md modified * api.md modified * YOLOv7 * yolov7 release link * yolov7 release link * yolov7 release link * copyright * change some helpers to private * change variables to const and fix documents. * gitignore * Transfer some funtions to private member of class * Transfer some funtions to private member of class * Merge from develop (#9) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * first commit for yolor * for merge * Develop (#11) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * Yolor (#16) * Develop (#11) (#12) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * Develop (#13) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * documents * documents * documents * documents * documents * documents * documents * documents * documents * documents * documents * documents * Develop (#14) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <928090362@qq.com> * add is_dynamic for YOLO series (#22) * first commit test photo * yolov7 doc * yolov7 doc * yolov7 doc * yolov7 doc * add yolov5 docs * modify yolov5 doc * first commit for retinaface * first commit for retinaface * firt commit for ultraface * firt commit for ultraface * firt commit for yolov5face * firt commit for modnet and arcface * firt commit for modnet and arcface * first commit for partial_fc * first commit for partial_fc * first commit for yolox * first commit for yolov6 * first commit for nano_det Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <928090362@qq.com>
This commit is contained in:
ziqi-jin
22
examples/vision/detection/nanodet_plus/README.md
Normal file
22
examples/vision/detection/nanodet_plus/README.md
Normal file
@@ -0,0 +1,22 @@
|
|||||||
|
# NanoDetPlus准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [NanoDetPlus v1.0.0-alpha-1](https://github.com/RangiLyu/nanodet/releases/tag/v1.0.0-alpha-1)
|
||||||
|
- （1）[链接中](https://github.com/RangiLyu/nanodet/releases/tag/v1.0.0-alpha-1)的*.onnx可直接进行部署
|
||||||
|
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了NanoDetPlus导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [NanoDetPlus_320](https://bj.bcebos.com/paddlehub/fastdeploy/nanodet-plus-m_320.onnx ) | 4.6MB | 27.0% |
|
||||||
|
| [NanoDetPlus_320_sim](https://bj.bcebos.com/paddlehub/fastdeploy/nanodet-plus-m_320-sim.onnx) | 4.6MB | 27.0% |
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/detection/nanodet_plus/cpp/CMakeLists.txt
Normal file
14
examples/vision/detection/nanodet_plus/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/detection/nanodet_plus/cpp/README.md
Normal file
85
examples/vision/detection/nanodet_plus/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# NanoDetPlus C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成NanoDetPlus在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的NanoDetPlus模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/nanodet-plus-m_320.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo nanodet-plus-m_320.onnx 000000014439.jpg 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo nanodet-plus-m_320.onnx 000000014439.jpg 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo nanodet-plus-m_320.onnx 000000014439.jpg 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## NanoDetPlus C++接口
|
||||||
|
|
||||||
|
### NanoDetPlus类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::detection::NanoDetPlus(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
NanoDetPlus模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> NanoDetPlus::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/detection/nanodet_plus/python/README.md
Normal file
79
examples/vision/detection/nanodet_plus/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# NanoDetPlus Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成NanoDetPlus在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载NanoDetPlus模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/nanodet-plus-m_320.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/nanodet_plus/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model nanodet-plus-m_320.onnx --image 000000014439.jpg --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model nanodet-plus-m_320.onnx --image 000000014439.jpg --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model nanodet-plus-m_320.onnx --image 000000014439.jpg --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## NanoDetPlus Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.detection.NanoDetPlus(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
NanoDetPlus模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> NanoDetPlus.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [NanoDetPlus 模型介绍](..)
|
||||||
|
- [NanoDetPlus C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
28
examples/vision/detection/yolov5/README.md
Normal file
28
examples/vision/detection/yolov5/README.md
Normal file
@@ -0,0 +1,28 @@
|
|||||||
|
# YOLOv7准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [YOLOv5 v6.0](https://github.com/ultralytics/yolov5/releases/tag/v6.0)
|
||||||
|
- （1）[链接中](https://github.com/ultralytics/yolov5/releases/tag/v6.0)的*.onnx可直接进行部署；
|
||||||
|
- （2）开发者基于自己数据训练的YOLOv5 v6.0模型，可使用[YOLOv5](https://github.com/ultralytics/yolov5)中的`export.py`导出ONNX文件后后，完成部署。
|
||||||
|
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了YOLOv7导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [YOLOv5n](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5n.onnx) | 1.9MB | 28.4% |
|
||||||
|
| [YOLOv5s](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5s.onnx) | 7.2MB | 37.2% |
|
||||||
|
| [YOLOv5m](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5m.onnx) | 21.2MB | 45.2% |
|
||||||
|
| [YOLOv5l](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5l.onnx) | 46.5MB | 48.8% |
|
||||||
|
| [YOLOv5x](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5x.onnx) | 86.7MB | 50.7% |
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/detection/yolov5/cpp/CMakeLists.txt
Normal file
14
examples/vision/detection/yolov5/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/detection/yolov5/cpp/README.md
Normal file
85
examples/vision/detection/yolov5/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# YOLOv5 C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成YOLOv5在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的yolov5模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov5s.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo yolov5s.onnx 000000014439.jpg 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo yolov5s.onnx 000000014439.jpg 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo yolov5s.onnx 000000014439.jpg 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOv5 C++接口
|
||||||
|
|
||||||
|
### YOLOv5类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::detection::YOLOv5(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOv5模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOv5::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
105
examples/vision/detection/yolov5/cpp/infer.cc
Normal file
105
examples/vision/detection/yolov5/cpp/infer.cc
Normal file
@@ -0,0 +1,105 @@
|
|||||||
|
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
|
||||||
|
//
|
||||||
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
||||||
|
// you may not use this file except in compliance with the License.
|
||||||
|
// You may obtain a copy of the License at
|
||||||
|
//
|
||||||
|
// http://www.apache.org/licenses/LICENSE-2.0
|
||||||
|
//
|
||||||
|
// Unless required by applicable law or agreed to in writing, software
|
||||||
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
||||||
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||||
|
// See the License for the specific language governing permissions and
|
||||||
|
// limitations under the License.
|
||||||
|
|
||||||
|
#include "fastdeploy/vision.h"
|
||||||
|
|
||||||
|
void CpuInfer(const std::string& model_file, const std::string& image_file) {
|
||||||
|
auto model = fastdeploy::vision::detection::YOLOv5(model_file);
|
||||||
|
if (!model.Initialized()) {
|
||||||
|
std::cerr << "Failed to initialize." << std::endl;
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
|
auto im = cv::imread(image_file);
|
||||||
|
auto im_bak = im.clone();
|
||||||
|
|
||||||
|
fastdeploy::vision::DetectionResult res;
|
||||||
|
if (!model.Predict(&im, &res)) {
|
||||||
|
std::cerr << "Failed to predict." << std::endl;
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
|
auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res);
|
||||||
|
cv::imwrite("vis_result.jpg", vis_im);
|
||||||
|
std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
|
||||||
|
}
|
||||||
|
|
||||||
|
void GpuInfer(const std::string& model_file, const std::string& image_file) {
|
||||||
|
auto option = fastdeploy::RuntimeOption();
|
||||||
|
option.UseGpu();
|
||||||
|
auto model = fastdeploy::vision::detection::YOLOv5(model_file, "", option);
|
||||||
|
if (!model.Initialized()) {
|
||||||
|
std::cerr << "Failed to initialize." << std::endl;
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
|
auto im = cv::imread(image_file);
|
||||||
|
auto im_bak = im.clone();
|
||||||
|
|
||||||
|
fastdeploy::vision::DetectionResult res;
|
||||||
|
if (!model.Predict(&im, &res)) {
|
||||||
|
std::cerr << "Failed to predict." << std::endl;
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
|
auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res);
|
||||||
|
cv::imwrite("vis_result.jpg", vis_im);
|
||||||
|
std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
|
||||||
|
}
|
||||||
|
|
||||||
|
void TrtInfer(const std::string& model_file, const std::string& image_file) {
|
||||||
|
auto option = fastdeploy::RuntimeOption();
|
||||||
|
option.UseGpu();
|
||||||
|
option.UseTrtBackend();
|
||||||
|
option.SetTrtInputShape("images", {1, 3, 640, 640});
|
||||||
|
auto model = fastdeploy::vision::detection::YOLOv5(model_file, "", option);
|
||||||
|
if (!model.Initialized()) {
|
||||||
|
std::cerr << "Failed to initialize." << std::endl;
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
|
auto im = cv::imread(image_file);
|
||||||
|
auto im_bak = im.clone();
|
||||||
|
|
||||||
|
fastdeploy::vision::DetectionResult res;
|
||||||
|
if (!model.Predict(&im, &res)) {
|
||||||
|
std::cerr << "Failed to predict." << std::endl;
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
|
auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res);
|
||||||
|
cv::imwrite("vis_result.jpg", vis_im);
|
||||||
|
std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
|
||||||
|
}
|
||||||
|
|
||||||
|
int main(int argc, char* argv[]) {
|
||||||
|
if (argc < 4) {
|
||||||
|
std::cout << "Usage: infer_demo path/to/model path/to/image run_option, "
|
||||||
|
"e.g ./infer_model ./yolov5.onnx ./test.jpeg 0"
|
||||||
|
<< std::endl;
|
||||||
|
std::cout << "The data type of run_option is int, 0: run with cpu; 1: run "
|
||||||
|
"with gpu; 2: run with gpu and use tensorrt backend."
|
||||||
|
<< std::endl;
|
||||||
|
return -1;
|
||||||
|
}
|
||||||
|
|
||||||
|
if (std::atoi(argv[3]) == 0) {
|
||||||
|
CpuInfer(argv[1], argv[2]);
|
||||||
|
} else if (std::atoi(argv[3]) == 1) {
|
||||||
|
GpuInfer(argv[1], argv[2]);
|
||||||
|
} else if (std::atoi(argv[3]) == 2) {
|
||||||
|
TrtInfer(argv[1], argv[2]);
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
79
examples/vision/detection/yolov5/python/README.md
Normal file
79
examples/vision/detection/yolov5/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# YOLOv5 Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成YOLOv5在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载yolov5模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov5s.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/yolov5/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model yolov5s.onnx --image 000000014439.jpg --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model yolov5s.onnx --image 000000014439.jpg --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model yolov5s.onnx --image 000000014439.jpg --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOv5 Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.detection.YOLOv5(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOv5模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOv5.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [YOLOv5 模型介绍](..)
|
||||||
|
- [YOLOv5 C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
51
examples/vision/detection/yolov5/python/infer.py
Normal file
51
examples/vision/detection/yolov5/python/infer.py
Normal file
@@ -0,0 +1,51 @@
|
|||||||
|
import fastdeploy as fd
|
||||||
|
import cv2
|
||||||
|
|
||||||
|
|
||||||
|
def parse_arguments():
|
||||||
|
import argparse
|
||||||
|
import ast
|
||||||
|
parser = argparse.ArgumentParser()
|
||||||
|
parser.add_argument(
|
||||||
|
"--model", required=True, help="Path of yolov5 onnx model.")
|
||||||
|
parser.add_argument(
|
||||||
|
"--image", required=True, help="Path of test image file.")
|
||||||
|
parser.add_argument(
|
||||||
|
"--device",
|
||||||
|
type=str,
|
||||||
|
default='cpu',
|
||||||
|
help="Type of inference device, support 'cpu' or 'gpu'.")
|
||||||
|
parser.add_argument(
|
||||||
|
"--use_trt",
|
||||||
|
type=ast.literal_eval,
|
||||||
|
default=False,
|
||||||
|
help="Wether to use tensorrt.")
|
||||||
|
return parser.parse_args()
|
||||||
|
|
||||||
|
|
||||||
|
def build_option(args):
|
||||||
|
option = fd.RuntimeOption()
|
||||||
|
|
||||||
|
if args.device.lower() == "gpu":
|
||||||
|
option.use_gpu()
|
||||||
|
|
||||||
|
if args.use_trt:
|
||||||
|
option.use_trt_backend()
|
||||||
|
option.set_trt_input_shape("images", [1, 3, 640, 640])
|
||||||
|
return option
|
||||||
|
|
||||||
|
|
||||||
|
args = parse_arguments()
|
||||||
|
|
||||||
|
# 配置runtime，加载模型
|
||||||
|
runtime_option = build_option(args)
|
||||||
|
model = fd.vision.detection.YOLOv5(args.model, runtime_option=runtime_option)
|
||||||
|
|
||||||
|
# 预测图片检测结果
|
||||||
|
im = cv2.imread(args.image)
|
||||||
|
result = model.predict(im)
|
||||||
|
|
||||||
|
# 预测结果可视化
|
||||||
|
vis_im = fd.vision.vis_detection(im, result)
|
||||||
|
cv2.imwrite("visualized_result.jpg", vis_im)
|
||||||
|
print("Visualized result save in ./visualized_result.jpg")
|
||||||
23
examples/vision/detection/yolov6/README.md
Normal file
23
examples/vision/detection/yolov6/README.md
Normal file
@@ -0,0 +1,23 @@
|
|||||||
|
# YOLOv6准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [YOLOv6 v0.1.0](https://github.com/meituan/YOLOv6/releases/download/0.1.0)
|
||||||
|
- （1）[链接中](https://github.com/meituan/YOLOv6/releases/download/0.1.0)的*.onnx可直接进行部署；
|
||||||
|
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了YOLOv6导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [YOLOv6s](https://bj.bcebos.com/paddlehub/fastdeploy/yolov6s.onnx) | 66MB | 43.1% |
|
||||||
|
| [YOLOv6s_640](https://bj.bcebos.com/paddlehub/fastdeploy/yolov6s-640x640.onnx) | 66MB | 43.1% |
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/detection/yolov6/cpp/CMakeLists.txt
Normal file
14
examples/vision/detection/yolov6/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/detection/yolov6/cpp/README.md
Normal file
85
examples/vision/detection/yolov6/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# YOLOv6 C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成YOLOv6在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的YOLOv6模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov6s.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo yolov6s.onnx 000000014439.jpg 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo yolov6s.onnx 000000014439.jpg 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo yolov6s.onnx 000000014439.jpg 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOv6 C++接口
|
||||||
|
|
||||||
|
### YOLOv6类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::detection::YOLOv6(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOv6模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOv6::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/detection/yolov6/python/README.md
Normal file
79
examples/vision/detection/yolov6/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# YOLOv6 Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成YOLOv6在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载YOLOv6模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov6s.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/yolov6/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model yolov6s.onnx --image 000000014439.jpg --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model yolov6s.onnx --image 000000014439.jpg --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model yolov6s.onnx --image 000000014439.jpg --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOv6 Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.detection.YOLOv6(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOv6模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOv6.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [YOLOv6 模型介绍](..)
|
||||||
|
- [YOLOv6 C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
@@ -3,13 +3,14 @@
|
|||||||
## 模型版本说明
|
## 模型版本说明
|
||||||
|
|
||||||
- [YOLOv7 0.1](https://github.com/WongKinYiu/yolov7/releases/tag/v0.1)
|
- [YOLOv7 0.1](https://github.com/WongKinYiu/yolov7/releases/tag/v0.1)
|
||||||
- （1）[YOLOv7 0.1](https://github.com/WongKinYiu/yolov7/releases/tag/v0.1)链接中.pt后缀模型通过[导出ONNX模型](#导出ONNX模型)操作后，可直接部署；.onnx、.trt和 .pose后缀模型暂不支持部署；
|
- （1）[链接中](https://github.com/WongKinYiu/yolov7/releases/tag/v0.1)的*.pt通过[导出ONNX模型](#导出ONNX模型)操作后，可进行部署；
|
||||||
- （2）开发者基于自己数据训练的YOLOv7 0.1模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。
|
- （2）[链接中](https://github.com/WongKinYiu/yolov7/releases/tag/v0.1)的*.onnx、*.trt和 *.pose模型不支持部署；
|
||||||
|
- （3）开发者基于自己数据训练的YOLOv7 0.1模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。
|
||||||
|
|
||||||
## 导出ONNX模型
|
## 导出ONNX模型
|
||||||
|
|
||||||
```
|
```
|
||||||
# 下载yolov7模型文件，或准备训练好的YOLOv7模型文件
|
# 下载yolov7模型文件
|
||||||
wget https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7.pt
|
wget https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7.pt
|
||||||
|
|
||||||
# 导出onnx格式文件 (Tips: 对应 YOLOv7 release v0.1 代码)
|
# 导出onnx格式文件 (Tips: 对应 YOLOv7 release v0.1 代码)
|
||||||
@@ -18,18 +19,24 @@ python models/export.py --grid --dynamic --weights PATH/TO/yolov7.pt
|
|||||||
# 如果您的代码版本中有支持NMS的ONNX文件导出，请使用如下命令导出ONNX文件(请暂时不要使用 "--end2end"，我们后续将支持带有NMS的ONNX模型的部署)
|
# 如果您的代码版本中有支持NMS的ONNX文件导出，请使用如下命令导出ONNX文件(请暂时不要使用 "--end2end"，我们后续将支持带有NMS的ONNX模型的部署)
|
||||||
python models/export.py --grid --dynamic --weights PATH/TO/yolov7.pt
|
python models/export.py --grid --dynamic --weights PATH/TO/yolov7.pt
|
||||||
|
|
||||||
# 移动onnx文件到examples目录
|
# 移动onnx文件到demo目录
|
||||||
cp PATH/TO/yolov7.onnx PATH/TO/FastDeploy/examples/vision/detextion/yolov7/
|
cp PATH/TO/yolov7.onnx PATH/TO/model_zoo/vision/yolov7/
|
||||||
```
|
```
|
||||||
|
|
||||||
## 下载预训练模型
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
为了方便开发者的测试，下面提供了YOLOv7导出的各系列模型，开发者可直接下载使用。
|
为了方便开发者的测试，下面提供了YOLOv7导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
| 模型 | 大小 | 精度 |
|
| 模型 | 大小 | 精度 |
|
||||||
|:---------------------------------------------------------------- |:----- |:----- |
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
| [YOLOv7](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx) | 141MB | 51.4% |
|
| [YOLOv7](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx) | 141MB | 51.4% |
|
||||||
| [YOLOv7-x] | 10MB | 51.4% |
|
| [YOLOv7x](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7x.onnx) | 273MB | 53.1% |
|
||||||
|
| [YOLOv7-w6](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7-w6.onnx) | 269MB | 54.9% |
|
||||||
|
| [YOLOv7-e6](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7-e6.onnx) | 372MB | 56.0% |
|
||||||
|
| [YOLOv7-d6](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7-d6.onnx) | 511MB | 56.6% |
|
||||||
|
| [YOLOv7-e6e](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7-e6e.onnx) | 579MB | 56.8% |
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
## 详细部署文档
|
## 详细部署文档
|
||||||
|
|||||||
@@ -5,7 +5,7 @@
|
|||||||
在部署前，需确认以下两个步骤
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuilt_libraries.md)
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
@@ -19,17 +19,21 @@ make -j
|
|||||||
|
|
||||||
#下载官方转换好的yolov7模型文件和测试图片
|
#下载官方转换好的yolov7模型文件和测试图片
|
||||||
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx
|
||||||
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000087038.jpg
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
# CPU推理
|
# CPU推理
|
||||||
./infer_demo yolov7.onnx 000000087038.jpg 0
|
./infer_demo yolov7.onnx 000000014439.jpg 0
|
||||||
# GPU推理
|
# GPU推理
|
||||||
./infer_demo yolov7.onnx 000000087038.jpg 1
|
./infer_demo yolov7.onnx 000000014439.jpg 1
|
||||||
# GPU上TensorRT推理
|
# GPU上TensorRT推理
|
||||||
./infer_demo yolov7.onnx 000000087038.jpg 2
|
./infer_demo yolov7.onnx 000000014439.jpg 2
|
||||||
```
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
## YOLOv7 C++接口
|
## YOLOv7 C++接口
|
||||||
|
|
||||||
### YOLOv7类
|
### YOLOv7类
|
||||||
@@ -58,11 +62,11 @@ YOLOv7模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
|||||||
> float conf_threshold = 0.25,
|
> float conf_threshold = 0.25,
|
||||||
> float nms_iou_threshold = 0.5)
|
> float nms_iou_threshold = 0.5)
|
||||||
> ```
|
> ```
|
||||||
>
|
>
|
||||||
> 模型预测接口，输入图像直接输出检测结果。
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
>
|
>
|
||||||
> **参数**
|
> **参数**
|
||||||
>
|
>
|
||||||
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
> > * **conf_threshold**: 检测框置信度过滤阈值
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
@@ -70,7 +74,11 @@ YOLOv7模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
|||||||
|
|
||||||
### 类成员变量
|
### 类成员变量
|
||||||
|
|
||||||
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
- [模型介绍](../../)
|
- [模型介绍](../../)
|
||||||
- [Python部署](../python)
|
- [Python部署](../python)
|
||||||
|
|||||||
@@ -18,15 +18,17 @@ git clone https://github.com/PaddlePaddle/FastDeploy.git
|
|||||||
cd examples/vison/detection/yolov7/python/
|
cd examples/vison/detection/yolov7/python/
|
||||||
|
|
||||||
# CPU推理
|
# CPU推理
|
||||||
python infer.py --model yolov7.onnx --image 000000087038.jpg --device cpu
|
python infer.py --model yolov7.onnx --image 000000014439.jpg --device cpu
|
||||||
# GPU推理
|
# GPU推理
|
||||||
python infer.py --model yolov7.onnx --image 000000087038.jpg --device gpu
|
python infer.py --model yolov7.onnx --image 000000014439.jpg --device gpu
|
||||||
# GPU上使用TensorRT推理 （注意：TensorRT推理第一次运行，有序列化模型的操作，有一定耗时，需要耐心等待）
|
# GPU上使用TensorRT推理
|
||||||
python infer.py --model yolov7.onnx --image 000000087038.jpg --device gpu --use_trt True
|
python infer.py --model yolov7.onnx --image 000000014439.jpg --device gpu --use_trt True
|
||||||
```
|
```
|
||||||
|
|
||||||
运行完成可视化结果如下图所示
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
## YOLOv7 Python接口
|
## YOLOv7 Python接口
|
||||||
|
|
||||||
```
|
```
|
||||||
@@ -47,22 +49,28 @@ YOLOv7模型加载和初始化，其中model_file为导出的ONNX模型格式
|
|||||||
> ```
|
> ```
|
||||||
> YOLOv7.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
> YOLOv7.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
> ```
|
> ```
|
||||||
>
|
>
|
||||||
> 模型预测结口，输入图像直接输出检测结果。
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
>
|
>
|
||||||
> **参数**
|
> **参数**
|
||||||
>
|
>
|
||||||
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
> **返回**
|
> **返回**
|
||||||
>
|
>
|
||||||
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
### 类成员属性
|
### 类成员属性
|
||||||
|
|
||||||
> > * **size**(list | tuple): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
## 其它文档
|
## 其它文档
|
||||||
|
|
||||||
|
|||||||
23
examples/vision/detection/yolox/README.md
Normal file
23
examples/vision/detection/yolox/README.md
Normal file
@@ -0,0 +1,23 @@
|
|||||||
|
# YOLOX准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [YOLOX v0.1.1](https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0)
|
||||||
|
- （1）[链接中](https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0)的*.onnx可直接进行部署；
|
||||||
|
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了YOLOX导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [YOLOX-s](https://bj.bcebos.com/paddlehub/fastdeploy/yolox_s.onnx) | 35MB | 40.5% |
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/detection/yolox/cpp/CMakeLists.txt
Normal file
14
examples/vision/detection/yolox/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/detection/yolox/cpp/README.md
Normal file
85
examples/vision/detection/yolox/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# YOLOX C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成YOLOX在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的YOLOX模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolox_s.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo yolox_s.onnx 000000014439.jpg 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo yolox_s.onnx 000000014439.jpg 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo yolox_s.onnx 000000014439.jpg 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOX C++接口
|
||||||
|
|
||||||
|
### YOLOX类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::detection::YOLOX(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOX模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOX::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/detection/yolox/python/README.md
Normal file
79
examples/vision/detection/yolox/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# YOLOX Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成YOLOX在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载YOLOX模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolox_s.onnx
|
||||||
|
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/yolox/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model yolox_s.onnx --image 000000014439.jpg --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model yolox_s.onnx --image 000000014439.jpg --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model yolox_s.onnx --image 000000014439.jpg --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOX Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.detection.YOLOX(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOX模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOX.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [YOLOX 模型介绍](..)
|
||||||
|
- [YOLOX C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
54
examples/vision/facedet/retinaface/README.md
Normal file
54
examples/vision/facedet/retinaface/README.md
Normal file
@@ -0,0 +1,54 @@
|
|||||||
|
# RetinaFace准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [RetinaFace CommitID:b984b4b](https://github.com/biubug6/Pytorch_Retinaface/commit/b984b4b)
|
||||||
|
- （1）[链接中](https://github.com/biubug6/Pytorch_Retinaface/commit/b984b4b)的*.pt通过[导出ONNX模型](#导出ONNX模型)操作后，可进行部署；
|
||||||
|
- （2）开发者基于自己数据训练的RetinaFace CommitID:b984b4b模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。
|
||||||
|
|
||||||
|
## 导出ONNX模型
|
||||||
|
|
||||||
|
自动下载的模型文件是我们事先转换好的，如果您需要从RetinaFace官方repo导出ONNX，请参考以下步骤。
|
||||||
|
|
||||||
|
* 下载官方仓库并
|
||||||
|
```bash
|
||||||
|
git clone https://github.com/biubug6/Pytorch_Retinaface.git
|
||||||
|
```
|
||||||
|
* 下载预训练权重并放在weights文件夹
|
||||||
|
```text
|
||||||
|
./weights/
|
||||||
|
mobilenet0.25_Final.pth
|
||||||
|
mobilenetV1X0.25_pretrain.tar
|
||||||
|
Resnet50_Final.pth
|
||||||
|
```
|
||||||
|
* 运行convert_to_onnx.py导出ONNX模型文件
|
||||||
|
```bash
|
||||||
|
PYTHONPATH=. python convert_to_onnx.py --trained_model ./weights/mobilenet0.25_Final.pth --network mobile0.25 --long_side 640 --cpu
|
||||||
|
PYTHONPATH=. python convert_to_onnx.py --trained_model ./weights/Resnet50_Final.pth --network resnet50 --long_side 640 --cpu
|
||||||
|
```
|
||||||
|
注意：需要先对convert_to_onnx.py脚本中的--long_side参数增加类型约束，type=int.
|
||||||
|
* 使用onnxsim对模型进行简化
|
||||||
|
```bash
|
||||||
|
onnxsim FaceDetector.onnx Pytorch_RetinaFace_mobile0.25-640-640.onnx # mobilenet
|
||||||
|
onnxsim FaceDetector.onnx Pytorch_RetinaFace_resnet50-640-640.onnx # resnet50
|
||||||
|
```
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了RetinaFace导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [RetinaFace_mobile0.25-640](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-640-640.onnx) | 1.7MB | - |
|
||||||
|
| [RetinaFace_mobile0.25-720](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-720-1080.onnx) | 1.7MB | -|
|
||||||
|
| [RetinaFace_resnet50-640](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_resnet50-720-1080.onnx) | 105MB | - |
|
||||||
|
| [RetinaFace_resnet50-720](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_resnet50-640-640.onnx) | 105MB | - |
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/facedet/retinaface/cpp/CMakeLists.txt
Normal file
14
examples/vision/facedet/retinaface/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/facedet/retinaface/cpp/README.md
Normal file
85
examples/vision/facedet/retinaface/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# RetinaFace C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成RetinaFace在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的RetinaFace模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-640-640.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo Pytorch_RetinaFace_mobile0.25-640-640.onnx todo 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo Pytorch_RetinaFace_mobile0.25-640-640.onnx todo 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo Pytorch_RetinaFace_mobile0.25-640-640.onnx todo 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## RetinaFace C++接口
|
||||||
|
|
||||||
|
### RetinaFace类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::facedet::RetinaFace(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
RetinaFace模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> RetinaFace::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/facedet/retinaface/python/README.md
Normal file
79
examples/vision/facedet/retinaface/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# RetinaFace Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成RetinaFace在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载retinaface模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-640-640.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/retinaface/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model Pytorch_RetinaFace_mobile0.25-640-640.onnx --image todo --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model Pytorch_RetinaFace_mobile0.25-640-640.onnx --image todo --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model Pytorch_RetinaFace_mobile0.25-640-640.onnx --image todo --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## RetinaFace Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.facedet.RetinaFace(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
RetinaFace模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> RetinaFace.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [RetinaFace 模型介绍](..)
|
||||||
|
- [RetinaFace C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
23
examples/vision/facedet/ultraface/README.md
Normal file
23
examples/vision/facedet/ultraface/README.md
Normal file
@@ -0,0 +1,23 @@
|
|||||||
|
# UltraFace准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [UltraFace CommitID:dffdddd](https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB/commit/dffdddd)
|
||||||
|
- （1）[链接中](https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB/commit/dffdddd)的*.onnx可下载, 也可以通过下面模型链接下载并进行部署
|
||||||
|
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了UltraFace导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [RFB-320](https://bj.bcebos.com/paddlehub/fastdeploy/version-RFB-320.onnx) | 1.3MB | - |
|
||||||
|
| [RFB-320-sim](https://bj.bcebos.com/paddlehub/fastdeploy/version-RFB-320-sim.onnx) | 1.2MB | -|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/facedet/ultraface/cpp/CMakeLists.txt
Normal file
14
examples/vision/facedet/ultraface/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/facedet/ultraface/cpp/README.md
Normal file
85
examples/vision/facedet/ultraface/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# UltraFace C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成UltraFace在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的UltraFace模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/version-RFB-320.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo version-RFB-320.onnx todo 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo version-RFB-320.onnx todo 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo version-RFB-320.onnx todo 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## UltraFace C++接口
|
||||||
|
|
||||||
|
### UltraFace类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::facedet::UltraFace(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
UltraFace模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> UltraFace::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/facedet/ultraface/python/README.md
Normal file
79
examples/vision/facedet/ultraface/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# UltraFace Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成UltraFace在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载ultraface模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/version-RFB-320.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/ultraface/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model version-RFB-320.onnx --image todo --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model version-RFB-320.onnx --image todo --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model version-RFB-320.onnx --image todo --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## UltraFace Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.facedet.UltraFace(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
UltraFace模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> UltraFace.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [UltraFace 模型介绍](..)
|
||||||
|
- [UltraFace C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
42
examples/vision/facedet/yolov5face/README.md
Normal file
42
examples/vision/facedet/yolov5face/README.md
Normal file
@@ -0,0 +1,42 @@
|
|||||||
|
# YOLOv5Face准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [YOLOv5Face CommitID:4fd1ead](https://github.com/deepcam-cn/yolov5-face/commit/4fd1ead)
|
||||||
|
- （1）[链接中](https://github.com/deepcam-cn/yolov5-face/commit/4fd1ead)的*.pt通过[导出ONNX模型](#导出ONNX模型)操作后，可进行部署；
|
||||||
|
- （2）开发者基于自己数据训练的YOLOv5Face CommitID:b984b4b模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。
|
||||||
|
|
||||||
|
## 导出ONNX模型
|
||||||
|
|
||||||
|
访问[YOLOv5Face](https://github.com/deepcam-cn/yolov5-face)官方github库，按照指引下载安装，下载`yolov5s-face.pt` 模型，利用 `export.py` 得到`onnx`格式文件。
|
||||||
|
|
||||||
|
* 下载yolov5face模型文件
|
||||||
|
```
|
||||||
|
Link: https://pan.baidu.com/s/1fyzLxZYx7Ja1_PCIWRhxbw Link: eq0q
|
||||||
|
https://drive.google.com/file/d/1zxaHeLDyID9YU4-hqK7KNepXIwbTkRIO/view?usp=sharing
|
||||||
|
```
|
||||||
|
|
||||||
|
* 导出onnx格式文件
|
||||||
|
```bash
|
||||||
|
PYTHONPATH=. python export.py --weights weights/yolov5s-face.pt --img_size 640 640 --batch_size 1
|
||||||
|
```
|
||||||
|
* onnx模型简化(可选)
|
||||||
|
```bash
|
||||||
|
onnxsim yolov5s-face.onnx yolov5s-face.onnx
|
||||||
|
```
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了YOLOv5Face导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [YOLOv5s-Face](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5s-face.onnx) | 30MB | - |
|
||||||
|
| [YOLOv5s-Face-bak](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5face-s-640x640.bak.onnx) | 30MB | -|
|
||||||
|
| [YOLOv5l-Face](https://bj.bcebos.com/paddlehub/fastdeploy/yolov5face-l-640x640.onnx ) | 181MB | - |
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/facedet/yolov5face/cpp/CMakeLists.txt
Normal file
14
examples/vision/facedet/yolov5face/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/facedet/yolov5face/cpp/README.md
Normal file
85
examples/vision/facedet/yolov5face/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# YOLOv5Face C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成YOLOv5Face在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的YOLOv5Face模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov5s-face.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo yolov5s-face.onnx todo 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo yolov5s-face.onnx todo 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo yolov5s-face.onnx todo 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOv5Face C++接口
|
||||||
|
|
||||||
|
### YOLOv5Face类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::facedet::YOLOv5Face(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOv5Face模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOv5Face::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/facedet/yolov5face/python/README.md
Normal file
79
examples/vision/facedet/yolov5face/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# YOLOv5Face Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成YOLOv5Face在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载YOLOv5Face模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov5s-face.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/yolov5face/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model yolov5s-face.onnx --image todo --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model yolov5s-face.onnx --image todo --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model yolov5s-face.onnx --image todo --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## YOLOv5Face Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.facedet.YOLOv5Face(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
YOLOv5Face模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> YOLOv5Face.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [YOLOv5Face 模型介绍](..)
|
||||||
|
- [YOLOv5Face C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
40
examples/vision/faceid/arcface/README.md
Normal file
40
examples/vision/faceid/arcface/README.md
Normal file
@@ -0,0 +1,40 @@
|
|||||||
|
# ArcFace准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [ArcFace CommitID:babb9a5](https://github.com/deepinsight/insightface/commit/babb9a5)
|
||||||
|
- （1）[链接中](https://github.com/deepinsight/insightface/commit/babb9a5)的*.pt通过[导出ONNX模型](#导出ONNX模型)操作后，可进行部署；
|
||||||
|
- （2）开发者基于自己数据训练的ArcFace CommitID:babb9a5模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。
|
||||||
|
|
||||||
|
## 导出ONNX模型
|
||||||
|
|
||||||
|
访问[ArcFace](https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch)官方github库，按照指引下载安装，下载pt模型文件，利用 `torch2onnx.py` 得到`onnx`格式文件。
|
||||||
|
|
||||||
|
* 下载ArcFace模型文件
|
||||||
|
```
|
||||||
|
Link: https://pan.baidu.com/share/init?surl=CL-l4zWqsI1oDuEEYVhj-g code: e8pw
|
||||||
|
```
|
||||||
|
|
||||||
|
* 导出onnx格式文件
|
||||||
|
```bash
|
||||||
|
PYTHONPATH=. python ./torch2onnx.py ms1mv3_arcface_r100_fp16/backbone.pth --output ms1mv3_arcface_r100.onnx --network r100 --simplify 1
|
||||||
|
```
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
<!-- 为了方便开发者的测试，下面提供了RetinaFace导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [RetinaFace_mobile0.25-640](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-640-640.onnx) | 1.7MB | - |
|
||||||
|
| [RetinaFace_mobile0.25-720](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_mobile0.25-720-1080.onnx) | 1.7MB | -|
|
||||||
|
| [RetinaFace_resnet50-640](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_resnet50-720-1080.onnx) | 105MB | - |
|
||||||
|
| [RetinaFace_resnet50-720](https://bj.bcebos.com/paddlehub/fastdeploy/Pytorch_RetinaFace_resnet50-640-640.onnx) | 105MB | - | -->
|
||||||
|
|
||||||
|
todo
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/faceid/arcface/cpp/CMakeLists.txt
Normal file
14
examples/vision/faceid/arcface/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/faceid/arcface/cpp/README.md
Normal file
85
examples/vision/faceid/arcface/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# ArcFace C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成ArcFace在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的ArcFace模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/ms1mv3_arcface_r34.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo ms1mv3_arcface_r34.onnx todo 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo ms1mv3_arcface_r34.onnx todo 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo ms1mv3_arcface_r34.onnx todo 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## ArcFace C++接口
|
||||||
|
|
||||||
|
### ArcFace类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::faceid::ArcFace(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
ArcFace模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> ArcFace::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/faceid/arcface/python/README.md
Normal file
79
examples/vision/faceid/arcface/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# ArcFace Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成ArcFace在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载arcface模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/ms1mv3_arcface_r34.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/arcface/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model ms1mv3_arcface_r34.onnx --image todo --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model ms1mv3_arcface_r34.onnx --image todo --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model ms1mv3_arcface_r34.onnx --image todo --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## ArcFace Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.faceid.ArcFace(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
ArcFace模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> ArcFace.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [ArcFace 模型介绍](..)
|
||||||
|
- [ArcFace C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
37
examples/vision/faceid/partial_fc/README.md
Normal file
37
examples/vision/faceid/partial_fc/README.md
Normal file
@@ -0,0 +1,37 @@
|
|||||||
|
<!-- # RetinaFace准备部署模型 -->
|
||||||
|
|
||||||
|
<!-- ## 模型版本说明
|
||||||
|
|
||||||
|
- [ArcFace CommitID:babb9a5](https://github.com/deepinsight/insightface/commit/babb9a5)
|
||||||
|
- （1）[链接中](https://github.com/deepinsight/insightface/commit/babb9a5)的*.pt通过[导出ONNX模型](#导出ONNX模型)操作后，可进行部署；
|
||||||
|
- （2）开发者基于自己数据训练的RetinaFace CommitID:b984b4b模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。 -->
|
||||||
|
|
||||||
|
<!-- ## 导出ONNX模型
|
||||||
|
|
||||||
|
访问[ArcFace](https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch)官方github库，按照指引下载安装，下载pt模型文件，利用 `torch2onnx.py` 得到`onnx`格式文件。
|
||||||
|
|
||||||
|
* 下载ArcFace模型文件
|
||||||
|
```
|
||||||
|
Link: https://pan.baidu.com/share/init?surl=CL-l4zWqsI1oDuEEYVhj-g code: e8pw
|
||||||
|
```
|
||||||
|
|
||||||
|
* 导出onnx格式文件
|
||||||
|
```bash
|
||||||
|
PYTHONPATH=. python ./torch2onnx.py ms1mv3_arcface_r100_fp16/backbone.pth --output ms1mv3_arcface_r100.onnx --network r100 --simplify 1
|
||||||
|
``` -->
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了RetinaFace导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [partial_fc_glint360k_r50](https://bj.bcebos.com/paddlehub/fastdeploy/partial_fc_glint360k_r50.onnx) | 167MB | - |
|
||||||
|
| [partial_fc_glint360k_r100](https://bj.bcebos.com/paddlehub/fastdeploy/partial_fc_glint360k_r100.onnx) | 249MB | -|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/faceid/partial_fc/cpp/CMakeLists.txt
Normal file
14
examples/vision/faceid/partial_fc/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/faceid/partial_fc/cpp/README.md
Normal file
85
examples/vision/faceid/partial_fc/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# PartialFC C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成PartialFC在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的PartialFC模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/partial_fc_glint360k_r50.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo partial_fc_glint360k_r50.onnx todo 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo partial_fc_glint360k_r50.onnx todo 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo partial_fc_glint360k_r50.onnx todo 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## PartialFC C++接口
|
||||||
|
|
||||||
|
### PartialFC类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::faceid::PartialFC(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
PartialFC模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> PartialFC::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/faceid/partial_fc/python/README.md
Normal file
79
examples/vision/faceid/partial_fc/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# PartialFC Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成PartialFC在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载partial_fc模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/partial_fc_glint360k_r50.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/partial_fc/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model partial_fc_glint360k_r50.onnx --image todo --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model partial_fc_glint360k_r50.onnx --image todo --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model partial_fc_glint360k_r50.onnx --image todo --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## PartialFC Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.faceid.PartialFC(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
PartialFC模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> PartialFC.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [PartialFC 模型介绍](..)
|
||||||
|
- [PartialFC C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
42
examples/vision/matting/modnet/README.md
Normal file
42
examples/vision/matting/modnet/README.md
Normal file
@@ -0,0 +1,42 @@
|
|||||||
|
# MODNet准备部署模型
|
||||||
|
|
||||||
|
## 模型版本说明
|
||||||
|
|
||||||
|
- [MODNet CommitID:28165a4](https://github.com/ZHKKKe/MODNet/commit/28165a4)
|
||||||
|
- （1）[链接中](https://github.com/ZHKKKe/MODNet/commit/28165a4)的*.pt通过[导出ONNX模型](#导出ONNX模型)操作后，可进行部署；
|
||||||
|
- （2）开发者基于自己数据训练的MODNet CommitID:b984b4b模型，可按照[导出ONNX模型](#%E5%AF%BC%E5%87%BAONNX%E6%A8%A1%E5%9E%8B)后，完成部署。
|
||||||
|
|
||||||
|
## 导出ONNX模型
|
||||||
|
|
||||||
|
|
||||||
|
访问[MODNet](https://github.com/ZHKKKe/MODNet)官方github库，按照指引下载安装，下载模型文件，利用 `onnx/export_onnx.py` 得到`onnx`格式文件。
|
||||||
|
|
||||||
|
* 导出onnx格式文件
|
||||||
|
```bash
|
||||||
|
python -m onnx.export_onnx \
|
||||||
|
--ckpt-path=pretrained/modnet_photographic_portrait_matting.ckpt \
|
||||||
|
--output-path=pretrained/modnet_photographic_portrait_matting.onnx
|
||||||
|
```
|
||||||
|
|
||||||
|
## 下载预训练ONNX模型
|
||||||
|
|
||||||
|
为了方便开发者的测试，下面提供了MODNet导出的各系列模型，开发者可直接下载使用。
|
||||||
|
|
||||||
|
| 模型 | 大小 | 精度 |
|
||||||
|
|:---------------------------------------------------------------- |:----- |:----- |
|
||||||
|
| [modnet_photographic](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_photographic__portrait_matting.onnx) | 25MB | - |
|
||||||
|
| [modnet_webcam](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_webcam_portrait_matting.onnx) | 25MB | -|
|
||||||
|
| [modnet_photographic_256](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_photographic_portrait_matting-256x256.onnx) | 25MB | - |
|
||||||
|
| [modnet_webcam_256](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_webcam_portrait_matting-256x256.onnx) | 25MB | - |
|
||||||
|
| [modnet_photographic_512](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_photographic_portrait_matting-512x512.onnx) | 25MB | - |
|
||||||
|
| [modnet_webcam_512](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_webcam_portrait_matting-512x512.onnx) | 25MB | - |
|
||||||
|
| [modnet_photographic_1024](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_photographic_portrait_matting-1024x1024.onnx) | 25MB | - |
|
||||||
|
| [modnet_webcam_1024](https://bj.bcebos.com/paddlehub/fastdeploy/modnet_webcam_portrait_matting-1024x1024.onnx) | 25MB | -|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 详细部署文档
|
||||||
|
|
||||||
|
- [Python部署](python)
|
||||||
|
- [C++部署](cpp)
|
||||||
14
examples/vision/matting/modnet/cpp/CMakeLists.txt
Normal file
14
examples/vision/matting/modnet/cpp/CMakeLists.txt
Normal file
@@ -0,0 +1,14 @@
|
|||||||
|
PROJECT(infer_demo C CXX)
|
||||||
|
CMAKE_MINIMUM_REQUIRED (VERSION 3.12)
|
||||||
|
|
||||||
|
# 指定下载解压后的fastdeploy库路径
|
||||||
|
option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
|
||||||
|
|
||||||
|
include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
|
||||||
|
|
||||||
|
# 添加FastDeploy依赖头文件
|
||||||
|
include_directories(${FASTDEPLOY_INCS})
|
||||||
|
|
||||||
|
add_executable(infer_demo ${PROJECT_SOURCE_DIR}/infer.cc)
|
||||||
|
# 添加FastDeploy库依赖
|
||||||
|
target_link_libraries(infer_demo ${FASTDEPLOY_LIBS})
|
||||||
85
examples/vision/matting/modnet/cpp/README.md
Normal file
85
examples/vision/matting/modnet/cpp/README.md
Normal file
@@ -0,0 +1,85 @@
|
|||||||
|
# MODNet C++部署示例
|
||||||
|
|
||||||
|
本目录下提供`infer.cc`快速完成MODNet在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/compile/prebuild_libraries.md)
|
||||||
|
|
||||||
|
以Linux上CPU推理为例，在本目录执行如下命令即可完成编译测试
|
||||||
|
|
||||||
|
```
|
||||||
|
mkdir build
|
||||||
|
cd build
|
||||||
|
wget https://xxx.tgz
|
||||||
|
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||||
|
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||||
|
make -j
|
||||||
|
|
||||||
|
#下载官方转换好的MODNet模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/modnet_photographic__portrait_matting.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
./infer_demo modnet_photographic__portrait_matting.onnx todo 0
|
||||||
|
# GPU推理
|
||||||
|
./infer_demo modnet_photographic__portrait_matting.onnx todo 1
|
||||||
|
# GPU上TensorRT推理
|
||||||
|
./infer_demo modnet_photographic__portrait_matting.onnx todo 2
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## MODNet C++接口
|
||||||
|
|
||||||
|
### MODNet类
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy::vision::matting::MODNet(
|
||||||
|
const string& model_file,
|
||||||
|
const string& params_file = "",
|
||||||
|
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||||
|
const Frontend& model_format = Frontend::ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
MODNet模型加载和初始化，其中model_file为导出的ONNX模型格式。
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX时，此参数传入空字符串即可
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX格式
|
||||||
|
|
||||||
|
#### Predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> MODNet::Predict(cv::Mat* im, DetectionResult* result,
|
||||||
|
> float conf_threshold = 0.25,
|
||||||
|
> float nms_iou_threshold = 0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测接口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **im**: 输入图像，注意需为HWC，BGR格式
|
||||||
|
> > * **result**: 检测结果，包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
> > * **conf_threshold**: 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**: NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
### 类成员变量
|
||||||
|
|
||||||
|
> > * **size**(vector<int>): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(vector<float>): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=ture` 表示不使用填充的方式，默认值为`is_no_pad=false`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=false`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_pad`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
- [模型介绍](../../)
|
||||||
|
- [Python部署](../python)
|
||||||
|
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
79
examples/vision/matting/modnet/python/README.md
Normal file
79
examples/vision/matting/modnet/python/README.md
Normal file
@@ -0,0 +1,79 @@
|
|||||||
|
# MODNet Python部署示例
|
||||||
|
|
||||||
|
在部署前，需确认以下两个步骤
|
||||||
|
|
||||||
|
- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/quick_start/requirements.md)
|
||||||
|
- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../docs/quick_start/install.md)
|
||||||
|
|
||||||
|
本目录下提供`infer.py`快速完成MODNet在CPU/GPU，以及GPU上通过TensorRT加速部署的示例。执行如下脚本即可完成
|
||||||
|
|
||||||
|
```
|
||||||
|
#下载modnet模型文件和测试图片
|
||||||
|
wget https://bj.bcebos.com/paddlehub/fastdeploy/modnet_photographic__portrait_matting.onnx
|
||||||
|
wget todo
|
||||||
|
|
||||||
|
|
||||||
|
#下载部署示例代码
|
||||||
|
git clone https://github.com/PaddlePaddle/FastDeploy.git
|
||||||
|
cd examples/vison/detection/modnet/python/
|
||||||
|
|
||||||
|
# CPU推理
|
||||||
|
python infer.py --model modnet_photographic__portrait_matting.onnx --image todo --device cpu
|
||||||
|
# GPU推理
|
||||||
|
python infer.py --model modnet_photographic__portrait_matting.onnx --image todo --device gpu
|
||||||
|
# GPU上使用TensorRT推理
|
||||||
|
python infer.py --model modnet_photographic__portrait_matting.onnx --image todo --device gpu --use_trt True
|
||||||
|
```
|
||||||
|
|
||||||
|
运行完成可视化结果如下图所示
|
||||||
|
|
||||||
|
<img width="640" src="https://user-images.githubusercontent.com/67993288/183847558-abcd9a57-9cd9-4891-b09a-710963c99b74.jpg">
|
||||||
|
|
||||||
|
## MODNet Python接口
|
||||||
|
|
||||||
|
```
|
||||||
|
fastdeploy.vision.matting.MODNet(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||||
|
```
|
||||||
|
|
||||||
|
MODNet模型加载和初始化，其中model_file为导出的ONNX模型格式
|
||||||
|
|
||||||
|
**参数**
|
||||||
|
|
||||||
|
> * **model_file**(str): 模型文件路径
|
||||||
|
> * **params_file**(str): 参数文件路径，当模型格式为ONNX格式时，此参数无需设定
|
||||||
|
> * **runtime_option**(RuntimeOption): 后端推理配置，默认为None，即采用默认配置
|
||||||
|
> * **model_format**(Frontend): 模型格式，默认为ONNX
|
||||||
|
|
||||||
|
### predict函数
|
||||||
|
|
||||||
|
> ```
|
||||||
|
> MODNet.predict(image_data, conf_threshold=0.25, nms_iou_threshold=0.5)
|
||||||
|
> ```
|
||||||
|
>
|
||||||
|
> 模型预测结口，输入图像直接输出检测结果。
|
||||||
|
>
|
||||||
|
> **参数**
|
||||||
|
>
|
||||||
|
> > * **image_data**(np.ndarray): 输入数据，注意需为HWC，BGR格式
|
||||||
|
> > * **conf_threshold**(float): 检测框置信度过滤阈值
|
||||||
|
> > * **nms_iou_threshold**(float): NMS处理过程中iou阈值
|
||||||
|
|
||||||
|
> **返回**
|
||||||
|
>
|
||||||
|
> > 返回`fastdeploy.vision.DetectionResult`结构体，结构体说明参考文档[视觉模型预测结果](../../../../../docs/api/vision_results/)
|
||||||
|
|
||||||
|
### 类成员属性
|
||||||
|
|
||||||
|
> > * **size**(list[int]): 通过此参数修改预处理过程中resize的大小，包含两个整型元素，表示[width, height], 默认值为[640, 640]
|
||||||
|
> > * **padding_value**(list[float]): 通过此参数可以修改图片在resize时候做填充(padding)的值, 包含三个浮点型元素, 分别表示三个通道的值, 默认值为[114, 114, 114]
|
||||||
|
> > * **is_no_pad**(bool): 通过此参数让图片是否通过填充的方式进行resize, `is_no_pad=True` 表示不使用填充的方式，默认值为`is_no_pad=False`
|
||||||
|
> > * **is_mini_pad**(bool): 通过此参数可以将resize之后图像的宽高这是为最接近`size`成员变量的值, 并且满足填充的像素大小是可以被`stride`成员变量整除的。默认值为`is_mini_pad=False`
|
||||||
|
> > * **stride**(int): 配合`stris_mini_padide`成员变量使用, 默认值为`stride=32`
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 其它文档
|
||||||
|
|
||||||
|
- [MODNet 模型介绍](..)
|
||||||
|
- [MODNet C++部署](../cpp)
|
||||||
|
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
||||||
Reference in New Issue
Block a user