目标跟踪算法(bytetrack)-tensorrt部署教程

2024-06-22 03:36

本文主要是介绍目标跟踪算法(bytetrack)-tensorrt部署教程,希望对大家解决编程问题提供一定的参考价值,需要的开发者们随着小编来一起学习吧!

一、本机安装python环境

conda create -n bytetrace_env python=3.8
activate bytetrace_env
conda install pytorch torchvision cudatoolkit=10.1 -c

检测GPU是否可用,不可用不行

import torch
print(torch.cuda.is_available())

安装bytetrack

git clone https://github.com/ifzhang/ByteTrack.git
cd ByteTrack
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
python setup.py develop

在这里插入图片描述
上述即安装成功。
安装pycocotools

pip install cython
pip install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'

或者(Linux)

pip install git+https://gitee.com/pursuit_zhangyu/cocoapi.git#subdirectory=PythonAPI

windows下

pip install pycocotools-windows

二、安装tensorrt环境

下载tensorrt包
TensorRT-8.4.3.1.Windows10.x86_64.cuda-10.2.cudnn8.4.zip
在这里插入图片描述
将所有的dll复制到
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
并设置环境变量。
在这里插入图片描述
虚拟环境中python版本为python3.8

pip install tensorrt-8.4.3.1-cp38-none-win_amd64.whl

三、转换模型

https://pan.baidu.com/s/1PiP1kQfgxAIrnGUbFP6Wfg
qflm
获取bytetrack_s_mot17.pth.tar并创建pretrained进行存放

python tools/trt.py -f exps/example/mot/yolox_s_mix_det.py -c pretrained/bytetrack_s_mot17.pth.tar

最后在D:\git_clone\ByteTrack-main\YOLOX_outputs\yolox_s_mix_det目录下生成tensorrt模型与pth模型:
在这里插入图片描述

四、cmake生成eigen库并使用VS2015编译

https://pan.baidu.com/s/15kEfCxpy-T7tz60msxxExg
ueq4
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述

五、下载opencv450

https://nchc.dl.sourceforge.net/project/opencvlibrary/4.5.0/opencv-4.5.0-vc14_vc15.exe?viasf=1
安装D:\opencv450

六、cmake生成bytetrack并使用VS2015编译

修改CMakeLists.txt

cmake_minimum_required(VERSION 2.6)project(bytetrack)add_definitions(-std=c++11)option(CUDA_USE_STATIC_CUDA_RUNTIME OFF)
set(CMAKE_CXX_STANDARD 11)
set(CMAKE_BUILD_TYPE Debug)find_package(CUDA REQUIRED)include_directories(${PROJECT_SOURCE_DIR}/include)
include_directories(D:\VS2015_CUDA\TensorRT\eigen-3.3.9\build\install\include\eigen3)
link_directories(${PROJECT_SOURCE_DIR}/include)
# include and link dirs of cuda and tensorrt, you need adapt them if yours are different
# cuda
include_directories(C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include)
link_directories(C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\lib\x64)
# cudnn
include_directories(C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include)
link_directories(C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\lib\x64)
# tensorrt
include_directories(D:\VS2015_CUDA\TensorRT\TensorRT-8.4.3.1\include)
link_directories(D:\VS2015_CUDA\TensorRT\TensorRT-8.4.3.1\lib)set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall -D_MWAITXINTRIN_H_INCLUDED")set(OpenCV_INCLUDE_DIRS_DIRS D:\opencv450\build\include)
set(OpenCV_LIBS D:\opencv450\build\x64\vc14\lib)
include_directories(${OpenCV_INCLUDE_DIRS})file(GLOB My_Source_Files ${PROJECT_SOURCE_DIR}/src/*.cpp)
add_executable(bytetrack ${My_Source_Files})
target_link_libraries(bytetrack nvinfer)
target_link_libraries(bytetrack cudart)
target_link_libraries(bytetrack ${OpenCV_LIBS})add_definitions(-O2 -pthread)

使用CMake进行配置生成。
打开VS2015工程进行配置
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
opencv可以自己加进去编译。
在这里插入图片描述
修改bytetrack.cpp

int img_w = cap.get(cv::CAP_PROP_FRAME_WIDTH);
int img_h = cap.get(cv::CAP_PROP_FRAME_HEIGHT);

这里有个bug,高的参数不对,会影响后面的demo.mp4。
编译生成
在这里插入图片描述

七、使用Dependency Walker查看exe依赖

在这里插入图片描述
在这里插入图片描述

D:\VS2015_CUDA\TensorRT\cpp\model_trt.engine -i D:\VS2015_CUDA\TensorRT\cpp\palace.mp4

八、windows源码

bytetrack.cpp

#include <fstream>
#include <iostream>
#include <sstream>
#include <numeric>
#include <chrono>
#include <vector>
#include <opencv2/opencv.hpp>
#include "NvInfer.h"
#include "cuda_runtime_api.h"
#include "logging.h"
#include "BYTETracker.h"#define CHECK(status) \do\{\auto ret = (status);\if (ret != 0)\{\cerr << "Cuda failure: " << ret << endl;\abort();\}\} while (0)#define DEVICE 0  // GPU id
#define NMS_THRESH 0.7
#define BBOX_CONF_THRESH 0.1using namespace nvinfer1;// stuff we know about the network and the input/output blobs
static const int INPUT_W = 1088;
static const int INPUT_H = 608;
const char* INPUT_BLOB_NAME = "input_0";
const char* OUTPUT_BLOB_NAME = "output_0";
static Logger gLogger;Mat static_resize(Mat& img) {float r = min(INPUT_W / (img.cols*1.0), INPUT_H / (img.rows*1.0));// r = std::min(r, 1.0f);int unpad_w = r * img.cols;int unpad_h = r * img.rows;Mat re(unpad_h, unpad_w, CV_8UC3);resize(img, re, re.size());Mat out(INPUT_H, INPUT_W, CV_8UC3, Scalar(114, 114, 114));re.copyTo(out(Rect(0, 0, re.cols, re.rows)));return out;
}struct GridAndStride
{int grid0;int grid1;int stride;
};static void generate_grids_and_stride(const int target_w, const int target_h, vector<int>& strides, vector<GridAndStride>& grid_strides)
{for (auto stride : strides){GridAndStride GS;int num_grid_w = target_w / stride;int num_grid_h = target_h / stride;for (int g1 = 0; g1 < num_grid_h; g1++){for (int g0 = 0; g0 < num_grid_w; g0++){GS.grid0 = g0;GS.grid1 = g1;GS.stride = stride;grid_strides.push_back(GS);}}}
}static inline float intersection_area(const Object& a, const Object& b)
{Rect_<float> inter = a.rect & b.rect;return inter.area();
}static void qsort_descent_inplace(vector<Object>& faceobjects, int left, int right)
{int i = left;int j = right;float p = faceobjects[(left + right) / 2].prob;while (i <= j){while (faceobjects[i].prob > p)i++;while (faceobjects[j].prob < p)j--;if (i <= j){// swapswap(faceobjects[i], faceobjects[j]);i++;j--;}}#pragma omp parallel sections{#pragma omp section{if (left < j) qsort_descent_inplace(faceobjects, left, j);}#pragma omp section{if (i < right) qsort_descent_inplace(faceobjects, i, right);}}
}static void qsort_descent_inplace(vector<Object>& objects)
{if (objects.empty())return;qsort_descent_inplace(objects, 0, objects.size() - 1);
}static void nms_sorted_bboxes(const vector<Object>& faceobjects, vector<int>& picked, float nms_threshold)
{picked.clear();const int n = faceobjects.size();vector<float> areas(n);for (int i = 0; i < n; i++){areas[i] = faceobjects[i].rect.area();}for (int i = 0; i < n; i++){const Object& a = faceobjects[i];int keep = 1;for (int j = 0; j < (int)picked.size(); j++){const Object& b = faceobjects[picked[j]];// intersection over unionfloat inter_area = intersection_area(a, b);float union_area = areas[i] + areas[picked[j]] - inter_area;// float IoU = inter_area / union_areaif (inter_area / union_area > nms_threshold)keep = 0;}if (keep)picked.push_back(i);}
}static void generate_yolox_proposals(vector<GridAndStride> grid_strides, float* feat_blob, float prob_threshold, vector<Object>& objects)
{const int num_class = 1;const int num_anchors = grid_strides.size();for (int anchor_idx = 0; anchor_idx < num_anchors; anchor_idx++){const int grid0 = grid_strides[anchor_idx].grid0;const int grid1 = grid_strides[anchor_idx].grid1;const int stride = grid_strides[anchor_idx].stride;const int basic_pos = anchor_idx * (num_class + 5);// yolox/models/yolo_head.py decode logicfloat x_center = (feat_blob[basic_pos+0] + grid0) * stride;float y_center = (feat_blob[basic_pos+1] + grid1) * stride;float w = exp(feat_blob[basic_pos+2]) * stride;float h = exp(feat_blob[basic_pos+3]) * stride;float x0 = x_center - w * 0.5f;float y0 = y_center - h * 0.5f;float box_objectness = feat_blob[basic_pos+4];for (int class_idx = 0; class_idx < num_class; class_idx++){float box_cls_score = feat_blob[basic_pos + 5 + class_idx];float box_prob = box_objectness * box_cls_score;if (box_prob > prob_threshold){Object obj;obj.rect.x = x0;obj.rect.y = y0;obj.rect.width = w;obj.rect.height = h;obj.label = class_idx;obj.prob = box_prob;objects.push_back(obj);}} // class loop} // point anchor loop
}float* blobFromImage(Mat& img){cvtColor(img, img, COLOR_BGR2RGB);float* blob = new float[img.total()*3];int channels = 3;int img_h = img.rows;int img_w = img.cols;vector<float> mean = {0.485f, 0.456f, 0.406f};vector<float> std = {0.229f, 0.224f, 0.225f};for (size_t c = 0; c < channels; c++) {for (size_t  h = 0; h < img_h; h++) {for (size_t w = 0; w < img_w; w++) {blob[c * img_w * img_h + h * img_w + w] =(((float)img.at<Vec3b>(h, w)[c]) / 255.0f - mean[c]) / std[c];}}}return blob;
}static void decode_outputs(float* prob, vector<Object>& objects, float scale, const int img_w, const int img_h) {vector<Object> proposals;vector<int> strides = {8, 16, 32};vector<GridAndStride> grid_strides;generate_grids_and_stride(INPUT_W, INPUT_H, strides, grid_strides);generate_yolox_proposals(grid_strides, prob,  BBOX_CONF_THRESH, proposals);//std::cout << "num of boxes before nms: " << proposals.size() << std::endl;qsort_descent_inplace(proposals);vector<int> picked;nms_sorted_bboxes(proposals, picked, NMS_THRESH);int count = picked.size();//std::cout << "num of boxes: " << count << std::endl;objects.resize(count);for (int i = 0; i < count; i++){objects[i] = proposals[picked[i]];// adjust offset to original unpaddedfloat x0 = (objects[i].rect.x) / scale;float y0 = (objects[i].rect.y) / scale;float x1 = (objects[i].rect.x + objects[i].rect.width) / scale;float y1 = (objects[i].rect.y + objects[i].rect.height) / scale;// clip// x0 = std::max(std::min(x0, (float)(img_w - 1)), 0.f);// y0 = std::max(std::min(y0, (float)(img_h - 1)), 0.f);// x1 = std::max(std::min(x1, (float)(img_w - 1)), 0.f);// y1 = std::max(std::min(y1, (float)(img_h - 1)), 0.f);objects[i].rect.x = x0;objects[i].rect.y = y0;objects[i].rect.width = x1 - x0;objects[i].rect.height = y1 - y0;}
}const float color_list[80][3] =
{{0.000, 0.447, 0.741},{0.850, 0.325, 0.098},{0.929, 0.694, 0.125},{0.494, 0.184, 0.556},{0.466, 0.674, 0.188},{0.301, 0.745, 0.933},{0.635, 0.078, 0.184},{0.300, 0.300, 0.300},{0.600, 0.600, 0.600},{1.000, 0.000, 0.000},{1.000, 0.500, 0.000},{0.749, 0.749, 0.000},{0.000, 1.000, 0.000},{0.000, 0.000, 1.000},{0.667, 0.000, 1.000},{0.333, 0.333, 0.000},{0.333, 0.667, 0.000},{0.333, 1.000, 0.000},{0.667, 0.333, 0.000},{0.667, 0.667, 0.000},{0.667, 1.000, 0.000},{1.000, 0.333, 0.000},{1.000, 0.667, 0.000},{1.000, 1.000, 0.000},{0.000, 0.333, 0.500},{0.000, 0.667, 0.500},{0.000, 1.000, 0.500},{0.333, 0.000, 0.500},{0.333, 0.333, 0.500},{0.333, 0.667, 0.500},{0.333, 1.000, 0.500},{0.667, 0.000, 0.500},{0.667, 0.333, 0.500},{0.667, 0.667, 0.500},{0.667, 1.000, 0.500},{1.000, 0.000, 0.500},{1.000, 0.333, 0.500},{1.000, 0.667, 0.500},{1.000, 1.000, 0.500},{0.000, 0.333, 1.000},{0.000, 0.667, 1.000},{0.000, 1.000, 1.000},{0.333, 0.000, 1.000},{0.333, 0.333, 1.000},{0.333, 0.667, 1.000},{0.333, 1.000, 1.000},{0.667, 0.000, 1.000},{0.667, 0.333, 1.000},{0.667, 0.667, 1.000},{0.667, 1.000, 1.000},{1.000, 0.000, 1.000},{1.000, 0.333, 1.000},{1.000, 0.667, 1.000},{0.333, 0.000, 0.000},{0.500, 0.000, 0.000},{0.667, 0.000, 0.000},{0.833, 0.000, 0.000},{1.000, 0.000, 0.000},{0.000, 0.167, 0.000},{0.000, 0.333, 0.000},{0.000, 0.500, 0.000},{0.000, 0.667, 0.000},{0.000, 0.833, 0.000},{0.000, 1.000, 0.000},{0.000, 0.000, 0.167},{0.000, 0.000, 0.333},{0.000, 0.000, 0.500},{0.000, 0.000, 0.667},{0.000, 0.000, 0.833},{0.000, 0.000, 1.000},{0.000, 0.000, 0.000},{0.143, 0.143, 0.143},{0.286, 0.286, 0.286},{0.429, 0.429, 0.429},{0.571, 0.571, 0.571},{0.714, 0.714, 0.714},{0.857, 0.857, 0.857},{0.000, 0.447, 0.741},{0.314, 0.717, 0.741},{0.50, 0.5, 0}
};void doInference(IExecutionContext& context, float* input, float* output, const int output_size, Size input_shape) {const ICudaEngine& engine = context.getEngine();// Pointers to input and output device buffers to pass to engine.// Engine requires exactly IEngine::getNbBindings() number of buffers.assert(engine.getNbBindings() == 2);void* buffers[2];// In order to bind the buffers, we need to know the names of the input and output tensors.// Note that indices are guaranteed to be less than IEngine::getNbBindings()const int inputIndex = engine.getBindingIndex(INPUT_BLOB_NAME);assert(engine.getBindingDataType(inputIndex) == nvinfer1::DataType::kFLOAT);const int outputIndex = engine.getBindingIndex(OUTPUT_BLOB_NAME);assert(engine.getBindingDataType(outputIndex) == nvinfer1::DataType::kFLOAT);int mBatchSize = engine.getMaxBatchSize();// Create GPU buffers on deviceCHECK(cudaMalloc(&buffers[inputIndex], 3 * input_shape.height * input_shape.width * sizeof(float)));CHECK(cudaMalloc(&buffers[outputIndex], output_size*sizeof(float)));// Create streamcudaStream_t stream;CHECK(cudaStreamCreate(&stream));// DMA input batch data to device, infer on the batch asynchronously, and DMA output back to hostCHECK(cudaMemcpyAsync(buffers[inputIndex], input, 3 * input_shape.height * input_shape.width * sizeof(float), cudaMemcpyHostToDevice, stream));context.enqueue(1, buffers, stream, nullptr);CHECK(cudaMemcpyAsync(output, buffers[outputIndex], output_size * sizeof(float), cudaMemcpyDeviceToHost, stream));cudaStreamSynchronize(stream);// Release stream and bufferscudaStreamDestroy(stream);CHECK(cudaFree(buffers[inputIndex]));CHECK(cudaFree(buffers[outputIndex]));
}int main(int argc, char** argv) {cudaSetDevice(DEVICE);// create a model using the API directly and serialize it to a streamchar *trtModelStream{nullptr};size_t size{0};if (argc == 4 && string(argv[2]) == "-i") {const string engine_file_path {argv[1]};ifstream file(engine_file_path, ios::binary);if (file.good()) {file.seekg(0, file.end);size = file.tellg();file.seekg(0, file.beg);trtModelStream = new char[size];assert(trtModelStream);file.read(trtModelStream, size);file.close();}} else {cerr << "arguments not right!" &

这篇关于目标跟踪算法(bytetrack)-tensorrt部署教程的文章就介绍到这儿,希望我们推荐的文章对编程师们有所帮助!



http://www.chinasem.cn/article/1083181

相关文章

闲置电脑也能活出第二春?鲁大师AiNAS让你动动手指就能轻松部署

对于大多数人而言,在这个“数据爆炸”的时代或多或少都遇到过存储告急的情况,这使得“存储焦虑”不再是个别现象,而将会是随着软件的不断臃肿而越来越普遍的情况。从不少手机厂商都开始将存储上限提升至1TB可以见得,我们似乎正处在互联网信息飞速增长的阶段,对于存储的需求也将会不断扩大。对于苹果用户而言,这一问题愈发严峻,毕竟512GB和1TB版本的iPhone可不是人人都消费得起的,因此成熟的外置存储方案开

Spring Security 从入门到进阶系列教程

Spring Security 入门系列 《保护 Web 应用的安全》 《Spring-Security-入门(一):登录与退出》 《Spring-Security-入门(二):基于数据库验证》 《Spring-Security-入门(三):密码加密》 《Spring-Security-入门(四):自定义-Filter》 《Spring-Security-入门(五):在 Sprin

不懂推荐算法也能设计推荐系统

本文以商业化应用推荐为例,告诉我们不懂推荐算法的产品,也能从产品侧出发, 设计出一款不错的推荐系统。 相信很多新手产品,看到算法二字,多是懵圈的。 什么排序算法、最短路径等都是相对传统的算法(注:传统是指科班出身的产品都会接触过)。但对于推荐算法,多数产品对着网上搜到的资源,都会无从下手。特别当某些推荐算法 和 “AI”扯上关系后,更是加大了理解的难度。 但,不了解推荐算法,就无法做推荐系

Makefile简明使用教程

文章目录 规则makefile文件的基本语法:加在命令前的特殊符号:.PHONY伪目标: Makefilev1 直观写法v2 加上中间过程v3 伪目标v4 变量 make 选项-f-n-C Make 是一种流行的构建工具,常用于将源代码转换成可执行文件或者其他形式的输出文件(如库文件、文档等)。Make 可以自动化地执行编译、链接等一系列操作。 规则 makefile文件

康拓展开(hash算法中会用到)

康拓展开是一个全排列到一个自然数的双射(也就是某个全排列与某个自然数一一对应) 公式: X=a[n]*(n-1)!+a[n-1]*(n-2)!+...+a[i]*(i-1)!+...+a[1]*0! 其中,a[i]为整数,并且0<=a[i]<i,1<=i<=n。(a[i]在不同应用中的含义不同); 典型应用: 计算当前排列在所有由小到大全排列中的顺序,也就是说求当前排列是第

csu 1446 Problem J Modified LCS (扩展欧几里得算法的简单应用)

这是一道扩展欧几里得算法的简单应用题,这题是在湖南多校训练赛中队友ac的一道题,在比赛之后请教了队友,然后自己把它a掉 这也是自己独自做扩展欧几里得算法的题目 题意:把题意转变下就变成了:求d1*x - d2*y = f2 - f1的解,很明显用exgcd来解 下面介绍一下exgcd的一些知识点:求ax + by = c的解 一、首先求ax + by = gcd(a,b)的解 这个

综合安防管理平台LntonAIServer视频监控汇聚抖动检测算法优势

LntonAIServer视频质量诊断功能中的抖动检测是一个专门针对视频稳定性进行分析的功能。抖动通常是指视频帧之间的不必要运动,这种运动可能是由于摄像机的移动、传输中的错误或编解码问题导致的。抖动检测对于确保视频内容的平滑性和观看体验至关重要。 优势 1. 提高图像质量 - 清晰度提升:减少抖动,提高图像的清晰度和细节表现力,使得监控画面更加真实可信。 - 细节增强:在低光条件下,抖

【数据结构】——原来排序算法搞懂这些就行,轻松拿捏

前言:快速排序的实现最重要的是找基准值,下面让我们来了解如何实现找基准值 基准值的注释:在快排的过程中,每一次我们要取一个元素作为枢纽值,以这个数字来将序列划分为两部分。 在此我们采用三数取中法,也就是取左端、中间、右端三个数,然后进行排序,将中间数作为枢纽值。 快速排序实现主框架: //快速排序 void QuickSort(int* arr, int left, int rig

阿里开源语音识别SenseVoiceWindows环境部署

SenseVoice介绍 SenseVoice 专注于高精度多语言语音识别、情感辨识和音频事件检测多语言识别: 采用超过 40 万小时数据训练,支持超过 50 种语言,识别效果上优于 Whisper 模型。富文本识别:具备优秀的情感识别,能够在测试数据上达到和超过目前最佳情感识别模型的效果。支持声音事件检测能力,支持音乐、掌声、笑声、哭声、咳嗽、喷嚏等多种常见人机交互事件进行检测。高效推

poj 3974 and hdu 3068 最长回文串的O(n)解法(Manacher算法)

求一段字符串中的最长回文串。 因为数据量比较大,用原来的O(n^2)会爆。 小白上的O(n^2)解法代码:TLE啦~ #include<stdio.h>#include<string.h>const int Maxn = 1000000;char s[Maxn];int main(){char e[] = {"END"};while(scanf("%s", s) != EO