三种方式实现人车流统计（yolov5+opencv+deepsort+bytetrack+iou）

一、运行环境

1、项目运行环境如下

2、CPU配置

3、GPU配置

如果没有GPU yolov5目标检测时间会比较久

二、编程语言与使用库版本

项目编程语言使用c++，使用的第三方库，onnxruntime-linux-x64-1.12.1，opencv-4.6.0

opencv 官方地址Releases - OpenCV

opencv github地址https://github.com/opencv/opencv/tree/4.10.0

onnxruntime 官方地址https://onnxruntime.ai/

 onnxruntime github 地址GitHub - microsoft/onnxruntime: ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

三、 检测模型

 1、项目使用yolov5目标检测模型

yolov5s.pt模型下载 GitHub - ultralytics/yolov5: YOLOv5 in PyTorch > ONNX > CoreML > TFLite

 

 2、使用命令 转换格式

python export.py --weights yolov5s.pt --include torchscript onnx

3、 使用feature.onnx 为特征提取模型

四、编译脚本

1、项目使用cmake 编写

创建文件CMakeLists.txt

cmake_minimum_required(VERSION 3.5)

add_definitions(-DPROJECT_PATH="${CMAKE_SOURCE_DIR}") 

project(DeepSORT LANGUAGES CXX)

set(CMAKE_CXX_STANDARD 14)
set(CMAKE_CXX_STANDARD_REQUIRED ON)

set(ONNXRUNTIME_DIR ${CMAKE_SOURCE_DIR}/lib/onnxruntime-linux-x64-1.12.1)

set(OpenCV_DIR ${CMAKE_SOURCE_DIR}/lib/opencv-4.6.0/install/lib/cmake/opencv4)  # 填入OpenCVConfig.cmake

include_directories("${ONNXRUNTIME_DIR}/include")

find_package(OpenCV 4 REQUIRED )

#message(STATUS "OpenCV_INCLUDE_DIRS: ${OpenCV_INCLUDE_DIRS}") 

include_directories(
    ${OpenCV_INCLUDE_DIRS}
    ${CMAKE_SOURCE_DIR}/tracker/deepsort/include
    ${CMAKE_SOURCE_DIR}/tracker/bytetrack/include
    ${CMAKE_SOURCE_DIR}/detector/YOLOv5/include
	${CMAKE_SOURCE_DIR}/include/eigen3
	${CMAKE_SOURCE_DIR}/tracker/com/include
	${CMAKE_SOURCE_DIR}/tracker/iou/include

    )
	
add_executable(DeepSORT
    detector/YOLOv5/src/YOLOv5Detector.cpp
	
    tracker/deepsort/src