mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-17 06:00:59 +08:00
Add sample doc and example code (#83)
* modify yolov7 and visualize functions * Add a sample document and example code * Delete .README.md.swp * Update README.md * Update README.md * Update README.md
This commit is contained in:
8
docs/api/vision_results/README.md
Normal file
8
docs/api/vision_results/README.md
Normal file
@@ -0,0 +1,8 @@
|
||||
# 视觉模型预测结果说明
|
||||
|
||||
FastDeploy根据视觉模型的任务类型,定义了不同的结构体(`csrcs/fastdeploy/vision/common/result.h`)来表达模型预测结果,具体如下表所示
|
||||
|
||||
| 结构体 | 文档 | 说明 | 相应模型 |
|
||||
| :----- | :--- | :---- | :------- |
|
||||
| ClassificationResult | [C++/Python文档](./classificiation_result.md) | 图像分类返回结果 | ResNet50、MobileNetV3等 |
|
||||
| DetectionResult | [C++/Python文档](./detection_result.md) | 目标检测返回结果 | PPYOLOE、YOLOv7系列模型等 |
|
28
docs/api/vision_results/classification_result.md
Normal file
28
docs/api/vision_results/classification_result.md
Normal file
@@ -0,0 +1,28 @@
|
||||
# ClassifyResult 图像分类结果
|
||||
|
||||
ClassifyResult代码定义在`csrcs/fastdeploy/vision/common/result.h`中,用于表明图像的分类结果和置信度。
|
||||
|
||||
## C++ 结构体
|
||||
|
||||
`fastdeploy::vision::ClassifyResult`
|
||||
|
||||
```
|
||||
struct ClassifyResult {
|
||||
std::vector<int32_t> label_ids;
|
||||
std::vector<float> scores;
|
||||
void Clear();
|
||||
std::string Str();
|
||||
};
|
||||
```
|
||||
|
||||
- **label_ids**: 成员变量,表示单张图片的分类结果,其个数根据在使用分类模型时传入的topk决定,例如可以返回top 5的分类结果
|
||||
- **scores**: 成员变量,表示单张图片在相应分类结果上的置信度,其个数根据在使用分类模型时传入的topk决定,例如可以返回top 5的分类置信度
|
||||
- **Clear()**: 成员函数,用于清除结构体中存储的结果
|
||||
- **Str()**: 成员函数,将结构体中的信息以字符串形式输出(用于Debug)
|
||||
|
||||
## Python结构体
|
||||
|
||||
`fastdeploy.vision.ClassifyResult`
|
||||
|
||||
- **label_ids**(list of int): 成员变量,表示单张图片的分类结果,其个数根据在使用分类模型时传入的topk决定,例如可以返回top 5的分类结果
|
||||
- **scores**(list of float): 成员变量,表示单张图片在相应分类结果上的置信度,其个数根据在使用分类模型时传入的topk决定,例如可以返回top 5的分类置信度
|
31
docs/api/vision_results/detection_result.md
Normal file
31
docs/api/vision_results/detection_result.md
Normal file
@@ -0,0 +1,31 @@
|
||||
# DetectionResult 目标检测结果
|
||||
|
||||
DetectionResult代码定义在`csrcs/fastdeploy/vision/common/result.h`中,用于表明图像检测出来的目标框、目标类别和目标置信度。
|
||||
|
||||
## C++ 结构体
|
||||
|
||||
`fastdeploy::vision::DetectionResult`
|
||||
|
||||
```
|
||||
struct DetectionResult {
|
||||
std::vector<std::array<float, 4>> boxes;
|
||||
std::vector<float> scores;
|
||||
std::vector<int32_t> label_ids;
|
||||
void Clear();
|
||||
std::string Str();
|
||||
};
|
||||
```
|
||||
|
||||
- **boxes**: 成员变量,表示单张图片检测出来的所有目标框坐标,`boxes.size()`表示框的个数,每个框以4个float数值依次表示xmin, ymin, xmax, ymax, 即左上角和右下角坐标
|
||||
- **scores**: 成员变量,表示单张图片检测出来的所有目标置信度,其元素个数与`boxes.size()`一致
|
||||
- **label_ids**: 成员变量,表示单张图片检测出来的所有目标类别,其元素个数与`boxes.size()`一致
|
||||
- **Clear()**: 成员函数,用于清除结构体中存储的结果
|
||||
- **Str()**: 成员函数,将结构体中的信息以字符串形式输出(用于Debug)
|
||||
|
||||
## Python结构体
|
||||
|
||||
`fastdeploy.vision.DetectionResult`
|
||||
|
||||
- **boxes**(list of list(float)): 成员变量,表示单张图片检测出来的所有目标框坐标。boxes是一个list,其每个元素为一个长度为4的list, 表示为一个框,每个框以4个float数值依次表示xmin, ymin, xmax, ymax, 即左上角和右下角坐标
|
||||
- **scores**(list of float): 成员变量,表示单张图片检测出来的所有目标置信度
|
||||
- **label_ids(list of int): 成员变量,表示单张图片检测出来的所有目标类别
|
21
docs/quick_start/install.md
Normal file
21
docs/quick_start/install.md
Normal file
@@ -0,0 +1,21 @@
|
||||
# FastDeploy安装
|
||||
|
||||
## Python安装
|
||||
|
||||
首先安装FastDeploy的SDK管理工具
|
||||
|
||||
```
|
||||
pip install fastdeploy-python
|
||||
```
|
||||
|
||||
然后通过命令行管理工具,根据自己的硬件安装相应SDK
|
||||
|
||||
### 安装CPU Python SDK
|
||||
```
|
||||
fastdeploy --install cpu
|
||||
```
|
||||
|
||||
### 安装GPU Python SDK
|
||||
```
|
||||
fastdeploy --install gpu
|
||||
```
|
21
docs/quick_start/requirements.md
Normal file
21
docs/quick_start/requirements.md
Normal file
@@ -0,0 +1,21 @@
|
||||
# FastDeploy环境要求
|
||||
|
||||
## 系统平台
|
||||
|
||||
- Linux x64/aarch64
|
||||
- Windows 10 x64
|
||||
- Mac OSX (x86 10.0以上/ arm64 12.0以上)
|
||||
|
||||
## 硬件
|
||||
|
||||
- Intel CPU
|
||||
- Nvidia GPU
|
||||
|
||||
## 软件
|
||||
|
||||
- cmake >= 3.12
|
||||
- gcc/g++ >= 8.2
|
||||
- python >= 3.6
|
||||
- Visual Studio 2019 (Windows平台)
|
||||
- cuda >= 11.0
|
||||
- cudnn >= 8.0
|
24
new_examples/vision/detection/README.md
Normal file
24
new_examples/vision/detection/README.md
Normal file
@@ -0,0 +1,24 @@
|
||||
# 视觉模型部署
|
||||
|
||||
本目录下提供了各类视觉模型的部署,主要涵盖以下任务类型
|
||||
|
||||
| 任务类型 | 说明 | 预测结果结构体 |
|
||||
| :------- | :----| :------------- |
|
||||
| Detection | 目标检测,输入图像,检测图像中物体位置,并返回检测框坐标及类别和置信度 | DetectionResult |
|
||||
| Segmentation | 语义分割,输入图像,给出图像中每个像素的分类及置信度 | SegmentationResult |
|
||||
| Classification | 图像分类,输入图像,给出图像的分类结果和置信度 | ClassificationResult |
|
||||
|
||||
|
||||
## FastDeploy API设计
|
||||
|
||||
视觉模型具有较有统一任务范式,在设计API时(包括C++/Python),FastDeploy将视觉模型的部署拆分为三个步骤
|
||||
|
||||
- 模型加载
|
||||
- 图像预处理
|
||||
- 模型推理
|
||||
- 推理结果后处理
|
||||
|
||||
FastDeploy针对飞桨的视觉套件,以及外部热门模型,提供端到端的部署服务,用户只需准备模型,按以下步骤即可完成整个模型的部署
|
||||
|
||||
- 加载模型
|
||||
- 调用`predict`接口
|
37
new_examples/vision/detection/yolov7/README.md
Normal file
37
new_examples/vision/detection/yolov7/README.md
Normal file
@@ -0,0 +1,37 @@
|
||||
# YOLOv7部署
|
||||
|
||||
## 版本依赖
|
||||
|
||||
- [YOLOv7 0.1](https://github.com/WongKinYiu/yolov7/releases/tag/v0.1)
|
||||
|
||||
|
||||
## 导出ONNX模型
|
||||
|
||||
```
|
||||
#下载yolov7模型文件
|
||||
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)
|
||||
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)
|
||||
python export.py --grid --dynamic --weights PATH/TO/yolov7.pt
|
||||
|
||||
# 移动onnx文件到demo目录
|
||||
cp PATH/TO/yolov7.onnx PATH/TO/model_zoo/vision/yolov7/
|
||||
```
|
||||
|
||||
## 预训练模型集合
|
||||
|
||||
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 |
|
||||
| :--- | :--- | :--- |
|
||||
| [YOLOv7](https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx) | 141MB | 51.4% |
|
||||
| [YOLOv7-x] | 10MB | 51.4% |
|
||||
|
||||
|
||||
## 模型部署
|
||||
|
||||
- [Python部署](python)
|
||||
- [C++部署](cpp)
|
14
new_examples/vision/detection/yolov7/cpp/CMakeLists.txt
Normal file
14
new_examples/vision/detection/yolov7/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 ${FASTDEPLOY_LIBS})
|
65
new_examples/vision/detection/yolov7/cpp/README.md
Normal file
65
new_examples/vision/detection/yolov7/cpp/README.md
Normal file
@@ -0,0 +1,65 @@
|
||||
# YOLOv7 C++部署示例
|
||||
|
||||
本目录下提供`infer.cc`快速完成YOLOv7在CPU/GPU,以及GPU上通过TensorRT加速部署的示例。
|
||||
|
||||
在部署前,需确认以下两个步骤
|
||||
|
||||
- 1. 软硬件环境满足要求,参考[FastDeploy环境要求](../../docs/quick_start/requirements.md)
|
||||
- 2. 根据开发环境,下载预编译部署库,参考[FastDeploy预编译库](../../docs/compile/prebuild_libraries.md)
|
||||
|
||||
以Linux上CPU推理为例,在本目录执行如下命令即可完成编译测试
|
||||
|
||||
```
|
||||
mkdir build
|
||||
cd build
|
||||
wget https://xxx.tgz
|
||||
tar xvf fastdeploy-linux-x64-0.2.0.tgz
|
||||
cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-0.2.0
|
||||
make -j
|
||||
|
||||
./infer_demo yolov7.onnx 000001.jpg
|
||||
```
|
||||
|
||||
## YOLOv7 C++接口
|
||||
|
||||
### YOLOv7类
|
||||
```
|
||||
fastdeploy::vision::wongkinyiu::YOLOv7(
|
||||
const string& model_file,
|
||||
const string& params_file = "",
|
||||
const RuntimeOption& runtime_option = RuntimeOption(),
|
||||
const Frontend& model_format = Frontend::ONNX)
|
||||
```
|
||||
YOLOv7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
|
||||
> ```
|
||||
> YOLOv7::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]
|
||||
|
||||
|
||||
- [模型介绍](../../)
|
||||
- [Python部署](../python)
|
||||
- [视觉模型预测结果](../../../../../docs/api/vision_results/)
|
106
new_examples/vision/detection/yolov7/cpp/infer.cc
Normal file
106
new_examples/vision/detection/yolov7/cpp/infer.cc
Normal file
@@ -0,0 +1,106 @@
|
||||
// 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::YOLOv7(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() {
|
||||
auto option = fastdeploy::RuntimeOption();
|
||||
option.UseGpu();
|
||||
auto model = fastdeploy::vision::detection::YOLOv7(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() {
|
||||
auto option = fastdeploy::RuntimeOption();
|
||||
option.UseGpu();
|
||||
option.UseTrtBackend();
|
||||
option.SetTrtInputShape("image", {1, 3, 320, 320}, {1, 3, 640, 640},
|
||||
{1, 3, 1280, 1280});
|
||||
auto model = fastdeploy::vision::detection::YOLOv7(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 ./yolov7.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();
|
||||
} else if (std::atoi(argv[3]) == 1) {
|
||||
GpuInfer();
|
||||
} else if (std::atoi(argv[3]) == 2) {
|
||||
TrtInfer();
|
||||
}
|
||||
return 0;
|
||||
}
|
62
new_examples/vision/detection/yolov7/python/README.md
Normal file
62
new_examples/vision/detection/yolov7/python/README.md
Normal file
@@ -0,0 +1,62 @@
|
||||
# YOLOv7 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YOLOv7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yolov7模型文件和测试图片
|
||||
wget https://bj.bcebos.com/paddlehub/fastdeploy/yolov7.onnx
|
||||
wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000087038.jpg
|
||||
|
||||
python infer.py --model yolov7.onnx --image 000000087038.jpg --device cpu
|
||||
```
|
||||
|
||||
运行完成可视化结果如下图所示
|
||||
|
||||
|
||||
## YOLOv7 Python接口
|
||||
|
||||
```
|
||||
fastdeploy.vision.detection.YOLOv7(model_file, params_file=None, runtime_option=None, model_format=Frontend.ONNX)
|
||||
```
|
||||
|
||||
YOLOv7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
|
||||
|
||||
> ```
|
||||
> 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。
|
||||
>
|
||||
> **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 | 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]
|
||||
|
||||
|
||||
## 其它文档
|
||||
|
||||
- [YOLOv7 模型介绍](..)
|
||||
- [YOLOv7 C++部署](../cpp)
|
||||
- [模型预测结果说明](../../../../../docs/api/vision_results/)
|
52
new_examples/vision/detection/yolov7/python/infer.py
Normal file
52
new_examples/vision/detection/yolov7/python/infer.py
Normal file
@@ -0,0 +1,52 @@
|
||||
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 yolov7 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("image", [1, 3, 320, 320],
|
||||
[1, 3, 640, 640], [1, 3, 1280, 1280])
|
||||
return option
|
||||
|
||||
|
||||
args = parse_arguments()
|
||||
|
||||
# 配置runtime,加载模型
|
||||
runtime_option = build_option(args)
|
||||
model = fd.vision.detection.YOLOv7(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")
|
Reference in New Issue
Block a user