{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "## 1. PP-Vehicle简介\n", "PaddleDetection深入探索核心行业的高频场景，提供了车辆场景的开箱即用分析工具，支持图片/单镜头视频/多镜头视频/在线视频流多种输入方式，广泛应用于智慧交通、智慧城市、工业巡检等领域。支持服务器端部署及TensorRT加速，T4服务器上可达到实时！\n", "PP-Vehicle囊括四大交通场景核心功能：车牌识别、属性识别、车流量统计、违章检测。\n", "\n", "PP-Vehicle由飞桨官方出品，是基于PaddleDetection的车辆分析pipeline。\n", "更多关于PaddleDetection可以点击https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.5/deploy/pipeline 进行了解。\n", "\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 2. 模型效果及应用场景\n", "### 2.1 PP-Vehicle 模型效果：\n", "\n", "| 任务 | 端到端速度（ms）| 模型方案 | 模型体积 |\n", "| :---------: | :-------: | :------: |:------: |\n", "| 车辆检测（高精度） | 25.7ms | [多目标跟踪](https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_l_36e_ppvehicle.zip) | 182M | \n", "| 车辆检测（轻量级） | 13.2ms | [多目标跟踪](https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_s_36e_ppvehicle.zip) | 27M |\n", "| 车辆跟踪（高精度） | 40ms | [多目标跟踪](https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_l_36e_ppvehicle.zip) | 182M |\n", "| 车辆跟踪（轻量级） | 25ms | [多目标跟踪](https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_s_36e_ppvehicle.zip) | 27M |\n", "| 车牌识别 | 4.68ms | [车牌检测](https://bj.bcebos.com/v1/paddledet/models/pipeline/ch_PP-OCRv3_det_infer.tar.gz)
[车牌字符识别](https://bj.bcebos.com/v1/paddledet/models/pipeline/ch_PP-OCRv3_rec_infer.tar.gz) | 车牌检测：3.9M
车牌字符识别： 12M |\n", "| 车辆属性 | 7.31ms | [车辆属性](https://bj.bcebos.com/v1/paddledet/models/pipeline/vehicle_attribute_model.zip) | 7.2M |\n", "\n", "\n", "### 2.2 应用场景：\n", "| 功能 | 方案优势 | 示例图 |\n", "| ---------- | ------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------- |\n", "| **车牌识别** | 支持传统车牌和新能源绿色车牌

车牌识别采用长间隔采样识别与多次结果统计投票方式，算力消耗少，识别精度高，结果稳定性好。检测模型 hmean: 0.979; 识别模型 acc: 0.773 | $\"\"$ |\n", "| **车辆属性分析** | 支持多种车型、颜色类别识别

使用更强力的Backbone模型PP-HGNet、PP-LCNet,精度高、速度快。识别精度: 90.81 | $\"\"$ |\n", "| **违章检测** | 简单易用：一行命令即可实现违停检测，自定义设置区域

检测、跟踪效果好，可实现违停车辆车牌识别 | $\"\"$ |\n", "| **车流量计数** | 简单易用：一行命令即可开启功能，自定义出入位置

可提供目标跟踪轨迹显示，统计准确度高 | $\"\"$ |\n", "\n", "\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 3. 模型如何使用\n", "\n", "（在Jupyter Notebook上运行时需要加\"！\"，若是cd命令则需加\"%\"）" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "vscode": { "languageId": "plaintext" } }, "outputs": [], "source": [ "# 克隆PaddleDetection仓库\n", "%cd ~/work\n", "!git clone https://github.com/PaddlePaddle/PaddleDetection.git\n", "\n", "# 安装其他依赖\n", "%cd PaddleDetection\n", "!pip install -r requirements.txt\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "\n", "\n", "### 3.1 配置文件说明\n", "\n", "PP-Vehicle相关配置位于```deploy/pipeline/config/infer_cfg_ppvehicle.yml```中，存放模型路径，完成不同功能需要设置不同的任务类型\n", "\n", "功能及任务类型对应表单如下：\n", "\n", "| 输入类型 | 功能 | 任务类型 | 配置项 |\n", "|-------|-------|----------|-----|\n", "| 图片 | 属性识别 | 目标检测属性识别 | DET ATTR |\n", "| 单镜头视频 | 属性识别 | 多目标跟踪属性识别 | MOT ATTR |\n", "| 单镜头视频 | 车牌识别 | 多目标跟踪车牌识别 | MOT VEHICLEPLATE |\n", "\n", "\n", "例如基于视频输入的属性识别，任务类型包含多目标跟踪和属性识别，具体配置如下：" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "vscode": { "languageId": "plaintext" } }, "outputs": [], "source": [ "\n", "```\n", "crop_thresh: 0.5\n", "visual: True\n", "warmup_frame: 50\n", "\n", "MOT:\n", " model_dir: https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_l_36e_ppvehicle.zip\n", " tracker_config: deploy/pipeline/config/tracker_config.yml\n", " batch_size: 1\n", " enable: True\n", "\n", "VEHICLE_ATTR:\n", " model_dir: https://bj.bcebos.com/v1/paddledet/models/pipeline/vehicle_attribute_model.zip\n", " batch_size: 8\n", " color_threshold: 0.5\n", " type_threshold: 0.5\n", " enable: True\n", "```" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**注意：**\n", "\n", "- 如果用户需要实现不同任务，可以在配置文件对应enable选项设置为True。\n", "- 如果用户仅需要修改模型文件路径，可以在命令行中--config后面紧跟着 `-o MOT.model_dir=ppyoloe/` 进行修改即可，也可以手动修改配置文件中的相应模型路径，详细说明参考下方参数说明文档。\n", "\n", "### 3.2 预测部署" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "vscode": { "languageId": "plaintext" } }, "outputs": [], "source": [ "# 1. 直接使用默认配置或者examples中配置文件，或者直接在`infer_cfg_ppvehicle.yml`中修改配置：\n", "# 例：车辆检测，指定配置文件路径和测试图片\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/infer_cfg_ppvehicle.yml --image_file=test_image.jpg --device=gpu\n", "\n", "# 例：车辆车牌识别，指定配置文件路径和测试视频\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/examples/infer_cfg_vehicle_plate.yml --video_file=test_video.mp4 --device=gpu\n", "\n", "\n", "#2. 使用命令行进行功能开启，或者模型路径修改：\n", "# 例：车辆跟踪，指定配置文件路径和测试视频，命令行中开启MOT模型并修改模型路径，命令行中指定的模型路径优先级高于配置文件\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/infer_cfg_ppvehicle.yml -o MOT.enable=True MOT.model_dir=ppyoloe_infer/ --video_file=test_video.mp4 --device=gpu\n", "\n", "# 例：车辆违章分析，指定配置文件和测试视频，命令行中指定违停区域设置、违停时间判断。\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/examples/infer_cfg_illegal_parking.yml \\\n", " --video_file=../car_test.mov \\\n", " --device=gpu \\\n", " --draw_center_traj \\\n", " --illegal_parking_time=3 \\\n", " --region_type=custom \\\n", " --region_polygon 600 300 1300 300 1300 800 600 800\n", "\n", "\n", "#3. rtsp推拉流\n", "\n", "#对rtsp拉流的支持，使用--rtsp RTSP [RTSP ...]参数指定一路或者多路rtsp视频流，如果是多路地址中间用空格隔开。(或者video_file后面的视频地址直接更换为rtsp流地址)，示例如下：\n", "\n", "# 例：车辆属性识别，单路视频流\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/examples/infer_cfg_vehicle_attr.yml -o visual=False --rtsp rtsp://[YOUR_RTSP_SITE] --device=gpu\n", "\n", "# 例：车辆属性识别，多路视频流\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/examples/infer_cfg_vehicle_attr.yml -o visual=False --rtsp rtsp://[YOUR_RTSP_SITE1] rtsp://[YOUR_RTSP_SITE2] --device=gpu\n", "\n", "\n", "#视频结果推流rtsp\n", "#预测结果进行rtsp推流，使用--pushurl rtsp:[IP] 推流到IP地址端，PC端可以使用[VLC播放器](https://vlc.onl/)打开网络流进行播放，播放地址为 `rtsp:[IP]/videoname`。其中`videoname`是预测的视频文件名，如果视频来源是本地摄像头则`videoname`默认为`output`.\n", "```\n", "# 例：车辆属性识别，单路视频流，该示例播放地址为 rtsp://[YOUR_SERVER_IP]:8554/test_video\n", "!python deploy/pipeline/pipeline.py --config deploy/pipeline/config/examples/infer_cfg_vehicle_attr.yml -o visual=False --video_file=test_video.mp4 --device=gpu --pushurl rtsp://[YOUR_SERVER_IP]:8554\n", "\n", "#注：\n", "#1. rtsp推流服务基于 [rtsp-simple-server](https://github.com/aler9/rtsp-simple-server), 如使用推流功能请先开启该服务.\n", "#2. rtsp推流如果模型处理速度跟不上会出现很明显的卡顿现象，建议跟踪模型使用ppyoloe_s版本，即修改配置中跟踪模型mot_ppyoloe_l_36e_pipeline.zip替换为mot_ppyoloe_s_36e_pipeline.zip。" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "\n", "### 3.3 Jetson部署说明\n", "\n", "由于Jetson平台算力相比服务器有较大差距，有如下使用建议：\n", "\n", "1. 模型选择轻量级版本，特别是跟踪模型，推荐使用`ppyoloe_s: https://bj.bcebos.com/v1/paddledet/models/pipeline/mot_ppyoloe_s_36e_pipeline.zip`\n", "2. 开启跟踪跳帧功能，推荐使用2或者3: `skip_frame_num: 3`\n", "\n", "使用该推荐配置，在TX2平台上可以达到较高速率，经测试属性案例达到20fps。\n", "\n", "可以直接修改配置文件（推荐），也可以在命令行中修改（字段较长，不推荐）。\n", "\n", "### 参数说明\n", "\n", "| 参数 | 是否必须|含义 |\n", "|-------|-------|----------|\n", "| --config | Yes | 配置文件路径 |\n", "| -o | Option | 覆盖配置文件中对应的配置 |\n", "| --image_file | Option | 需要预测的图片 |\n", "| --image_dir | Option | 要预测的图片文件夹路径 |\n", "| --video_file | Option | 需要预测的视频，或者rtsp流地址（推荐使用rtsp参数） |\n", "| --rtsp | Option | rtsp视频流地址，支持一路或者多路同时输入 |\n", "| --camera_id | Option | 用来预测的摄像头ID，默认为-1(表示不使用摄像头预测，可设置为：0 - (摄像头数目-1) )，预测过程中在可视化界面按`q`退出输出预测结果到：output/output.mp4|\n", "| --device | Option | 运行时的设备，可选择`CPU/GPU/XPU`，默认为`CPU`|\n", "| --pushurl | Option| 对预测结果视频进行推流的地址，以rtsp://开头，该选项优先级高于视频结果本地存储，打开时不再另外存储本地预测结果视频|\n", "| --output_dir | Option|可视化结果保存的根目录，默认为output/|\n", "| --run_mode | Option |使用GPU时，默认为paddle, 可选（paddle/trt_fp32/trt_fp16/trt_int8）|\n", "| --enable_mkldnn | Option | CPU预测中是否开启MKLDNN加速，默认为False |\n", "| --cpu_threads | Option| 设置cpu线程数，默认为1 |\n", "| --trt_calib_mode | Option| TensorRT是否使用校准功能，默认为False。使用TensorRT的int8功能时，需设置为True，使用PaddleSlim量化后的模型时需要设置为False |\n", "| --do_entrance_counting | Option | 是否统计出入口流量，默认为False |\n", "| --draw_center_traj | Option | 是否绘制跟踪轨迹，默认为False |\n", "| --region_type | Option | 'horizontal'（默认值）、'vertical'：表示流量统计方向选择；'custom'：表示设置闯入区域 |\n", "| --region_polygon | Option | 设置闯入区域多边形多点的坐标，无默认值 |\n", "| --do_break_in_counting | Option | 此项表示做区域闯入检查 |\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 4. 方案介绍\n", "PP-Vehicle 整体方案如下图所示:\n", "\n", "

\n", "

\n", "

\n", "\n", "\n", "### 车辆检测\n", "- 采用PP-YOLOE L 作为目标检测模型\n", "- 详细文档参考[PP-YOLOE](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/configs/ppyoloe)和[检测跟踪文档](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/deploy/pipeline/docs/tutorials/ppvehicle_mot.md)\n", "\n", "### 车辆跟踪\n", "- 采用SDE方案完成车辆跟踪\n", "- 检测模型使用PP-YOLOE L(高精度)和S(轻量级)\n", "- 跟踪模块采用OC-SORT方案\n", "- 详细文档参考[OC-SORT](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/configs/mot/ocsort)和[检测跟踪文档](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/deploy/pipeline/docs/tutorials/ppvehicle_mot.md)\n", "\n", "### 属性识别\n", "- 使用PaddleClas提供的特色模型PP-LCNet，实现对车辆颜色及车型属性的识别。\n", "- 详细文档参考[属性识别](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/deploy/pipeline/docs/tutorials/ppvehicle_attribute.md)\n", "\n", "### 车牌识别\n", "- 使用PaddleOCR特色模型ch_PP-OCRv3_det+ch_PP-OCRv3_rec模型，识别车牌号码\n", "- 详细文档参考[车牌识别](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/deploy/pipeline/docs/tutorials/ppvehicle_plate.md)\n", "\n", "### 违章停车识别\n", "- 车辆跟踪模型使用高精度模型PP-YOLOE L，根据车辆的跟踪轨迹以及指定的违停区域判断是否违章停车，如果存在则展示违章停车车牌号。\n", "- 详细文档参考[违章停车识别](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/deploy/pipeline/docs/tutorials/ppvehicle_illegal_parking.md)" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.10.8" } }, "nbformat": 4, "nbformat_minor": 4 }