基于 yolov5 的智能人脸数据自动打标工具 可自定义人脸检测模型、可导出多种格式标签

<p align="center"> <a href="https://gitee.com/CV_Lab/face-labeling"> <img src="https://pycver.gitee.io/ows-pics/imgs/facelabeling_logo.png" alt="Simple Icons" > </a> <p align="center"> 基于YOLOv5的智能人脸数据标注工具，实现人脸数据标注自动化 </p> <p align="center"> 可自定义人脸检测模型、可导出多种格式标签，包括PASCAL VOC XML、MS COCO JSON、YOLO TXT </p> </p> <p align="center"> <a href="./CodeCheck.md"><img src="https://img.shields.io/badge/CodeCheck-passing-success" alt="code check" /></a> <a href="https://gitee.com/CV_Lab/face-labeling/releases/v0.2.2"><img src="https://img.shields.io/badge/Releases-v0.2.2-green" alt="Releases Version" /></a> <a href="https://github.com/ultralytics/yolov5"><img src="https://img.shields.io/badge/YOLOv5-v7.0-blue" alt="YOLOv5 Version" /></a> <a href="http://host.robots.ox.ac.uk/pascal/VOC/"><img src="https://img.shields.io/badge/PASCAL%20VOC-XML-blue" alt="PASCAL VOC XML" /></a> <a href="https://cocodataset.org/"><img src="https://img.shields.io/badge/MS%20COCO-JSON-brightgreen" alt="MS COCO JSON" /></a> <a href="https://pjreddie.com/darknet/"><img src="https://img.shields.io/badge/YOLO-TXT-orange" alt="YOLO TXT" /></a> <a href="https://gitee.com/CV_Lab/face-labeling/blob/master/LICENSE"><img src="https://img.shields.io/badge/License-GPL--3.0-blue" alt="License" /></a> </p> <p align="center"> <a href="https://github.com/pre-commit/pre-commit"><img src="https://img.shields.io/badge/checks-pre--commit-brightgreen" alt="pre-commit"></a> <a href="#"><img src="https://img.shields.io/badge/Python-3.8%2B-blue?logo=python" alt="Python Version" /></a> <a href='https://gitee.com/CV_Lab/face-labeling/stargazers'><img src='https://gitee.com/CV_Lab/face-labeling/badge/star.svg?theme=dark' alt='star'></img></a> <a href='https://gitee.com/CV_Lab/face-labeling/members'><img src='https://gitee.com/CV_Lab/face-labeling/badge/fork.svg?theme=dark' alt='fork'></img></a> </p> ## 🚀 作者简介 曾逸夫，从事人工智能研究与开发；主研领域：计算机视觉；[YOLOv8官方开源项目代码贡献人](https://github.com/ultralytics/ultralytics/graphs/contributors)；[YOLOv5官方开源项目代码贡献人](https://github.com/ultralytics/yolov5/graphs/contributors)；[YOLOv5 v6.1代码贡献人](https://github.com/ultralytics/yolov5/releases/tag/v6.1)；[YOLOv5 v6.2代码贡献人](https://github.com/ultralytics/yolov5/releases/tag/v6.2)；[YOLOv5 v7.0代码贡献人](https://github.com/ultralytics/yolov5/releases/tag/v7.0)；[Gradio官方开源项目代码贡献人](https://github.com/gradio-app/gradio/graphs/contributors) ❤️ Github：https://github.com/Zengyf-CVer 🔥 YOLOv8 官方开源项目PR ID： - Fix YOLOv8 Chinese introduction：https://github.com/ultralytics/ultralytics/pull/219 - Add codespell and upgrade some components：https://github.com/ultralytics/ultralytics/pull/599 - Add pycln：https://github.com/ultralytics/ultralytics/pull/7141 - Fix heatmap width and height parameter bug：https://github.com/ultralytics/ultralytics/pull/7368 - ultralytics 8.0.237 cv2.CAP_PROP fix and in_counts and out_counts displays：https://github.com/ultralytics/ultralytics/pull/7380 💡 Ultralytics v8.1.0 代码贡献链接： - https://github.com/ultralytics/ultralytics/releases/tag/v8.1.0 🔥 YOLOv5 官方开源项目PR ID： - Save \*.npy features on detect.py `--visualize`：https://github.com/ultralytics/yolov5/pull/5701 - Fix `detect.py --view-img` for non-ASCII paths：https://github.com/ultralytics/yolov5/pull/7093 - Fix Flask REST API：https://github.com/ultralytics/yolov5/pull/7210 - Add yesqa to precommit checks：https://github.com/ultralytics/yolov5/pull/7511 - Add mdformat to precommit checks and update other version：https://github.com/ultralytics/yolov5/pull/7529 - Add TensorRT dependencies：https://github.com/ultralytics/yolov5/pull/8553 - Add paddle tips：https://github.com/ultralytics/yolov5/pull/9502 💡 YOLOv5 v6.1 & v6.2 & v7.0 代码贡献链接： - https://github.com/ultralytics/yolov5/releases/tag/v6.1 - https://github.com/ultralytics/yolov5/releases/tag/v6.2 - https://github.com/ultralytics/yolov5/releases/tag/v7.0 🔥 Gradio 官方开源项目PR ID： - Create a color generator demo：https://github.com/gradio-app/gradio/pull/1872 🔥 Gradio 官方开源项目PR ID： - Create a color generator demo：https://github.com/gradio-app/gradio/pull/1872 <h2 align="center">🚀更新走势</h2> - `2022-07-20` **⚡ [Face Labeling v0.2.2 发行版](https://gitee.com/CV_Lab/face-labeling/releases/v0.2.2)正式上线** - `2022-06-12` **⚡ [Face Labeling v0.2 发行版](https://gitee.com/CV_Lab/face-labeling/releases/v0.2)正式上线** - `2022-04-25` **⚡ [Face Labeling v0.1 发行版](https://gitee.com/CV_Lab/face-labeling/releases/v0.1)正式上线** <h2 align="center">💎项目流程与用途</h2> ### 📌 项目整体流程与扩展应用 <div align="center" > <img src="https://pycver.gitee.io/ows-pics/imgs/face_labeling_work_flow.png"> </div> ### 📌 项目功能结构与信息流 <div align="center" > <img src="https://pycver.gitee.io/ows-pics/imgs/face_labeling_work_io.png"> </div> <h2 align="center">💡项目结构</h2> ``` . ├── face-labeling # 项目名称 │ ├── util # 工具包 │ │ ├── voc_xml.py # PASCAL VOC XML │ │ ├── coco_json.py # MS COCO JSON │ │ ├── yolo_txt.py # YOLO TXT │ │ ├── model_opt.py # 模型管理 │ │ ├── obj_opt.py # 目标管理 │ │ ├── path_opt.py # 路径管理 │ │ ├── log.py # 日志管理 │ │ └── time_format.py # 日期格式化 │ ├── data # 测试数据 │ │ └── imgs # 测试图片，来源于WIDER FACE Test │ ├── models # 模型Hub │ │ ├── readme.md # 模型Hub README │ │ ├── *.pt # PyTorch模型 │ │ └── *.onnx # ONNX模型 │ ├── face_labeling.py # 主运行文件 │ ├── LICENSE # 项目许可 │ ├── CodeCheck.md # 代码检查 │ ├── .gitignore # git忽略文件 │ ├── yolov5_widerface.md # 基于YOLOv5的人脸检测模型的构建 │ ├── yolov5_pytorch_gpu.md # YOLOv5 PyTorch GPU安装教程 │ ├── README.md # 项目说明 │ └── requirements.txt # 脚本依赖包 ``` <h2 align="center">🔥安装教程</h2> ### ✅ 第一步：安装Face Labeling 📌 创建conda环境 ```shell conda create -n facelabel python==3.8 conda activate facelabel # 进入环境 ``` 📌 克隆 ```shell git clone https://gitee.com/CV_Lab/face-labeling.git ``` ### ✅ 第二步：安装Face Labeling依赖 ```shell cd ./face-labeling conda activate facelabel # 进入环境 pip install -r requirements.txt -U ``` 📌 将人脸模型文件（.pt）放入`models` 目录中 ❗ 注意：yolov5默认采用pip安装PyTorch GPU版，如果采用官网安装**PyTorch GPU**版，参见[YOLOv5 PyTorch GPU安装教程](./yolov5_pytorch_gpu.md) ### ✅ 基于YOLOv5的人脸检测模型的构建 📌 **widerface-m人脸检测模型**是在[WIDER FACE](http://shuoyang1213.me/WIDERFACE/)数据集上，基于[YOLOv5 v6.1](https://github.com/ultralytics/yolov5)训练的，具体训练过程参见[yolov5_widerface.md](./yolov5_widerface.md) 📌 **darkface-m人脸检测模型**是在[DARK FACE](https://flyywh.github.io/CVPRW2019LowLight/)数据集上，基于[YOLOv5 v6.1](https://github.com/ultralytics/yolov5)训练的，具体训练过程参见[yolov5_darkface.md](./yolov5_darkface.md) ❤️ 本项目提供了以下人脸检测模型： | 模型名称 | 下载地址 | 模型大小 | 适用范围 | 适用设备 | | :---------: |