docs(readme.zh-cn.md): new

zjykzj · Jul 16, 2023 · 112b0ed · 112b0ed
1 parent 7bca27b
commit 112b0ed
Showing 1 changed file with 327 additions and 0 deletions.
diff --git a/README.zh-CN.md b/README.zh-CN.md
@@ -0,0 +1,327 @@
+<div align="right">
+  语言:
+    🇨🇳
+  <a title="英语" href="./README.md">🇺🇸</a>
+</div>
+
+<div align="center"><a title="" href="https://github.com/zjykzj/YOLOv2"><img align="center" src="./imgs/YOLOv2.png" alt=""></a></div>
+
+<p align="center">
+  «YOLOv2» 复现了论文 "YOLO9000: Better, Faster, Stronger"
+<br>
+<br>
+  <a href="https://github.com/RichardLitt/standard-readme"><img src="https://img.shields.io/badge/standard--readme-OK-green.svg?style=flat-square" alt=""></a>
+  <a href="https://conventionalcommits.org"><img src="https://img.shields.io/badge/Conventional%20Commits-1.0.0-yellow.svg" alt=""></a>
+  <a href="http://commitizen.github.io/cz-cli/"><img src="https://img.shields.io/badge/commitizen-friendly-brightgreen.svg" alt=""></a>
+</p>
+
+* 使用`VOC07+12 trainval`数据集进行训练，使用`VOC2007 Test`进行测试，输入大小为`416x416`。测试结果如下：
+
+<!-- <style type="text/css">
+.tg  {border-collapse:collapse;border-spacing:0;}
+.tg td{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
+  overflow:hidden;padding:10px 5px;word-break:normal;}
+.tg th{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
+  font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;}
+.tg .tg-zkss{background-color:#FFF;border-color:inherit;color:#333;text-align:center;vertical-align:top}
+.tg .tg-c3ow{border-color:inherit;text-align:center;vertical-align:top}
+.tg .tg-fr9f{background-color:#FFF;border-color:inherit;color:#333;font-weight:bold;text-align:center;vertical-align:top}
+.tg .tg-y5w1{background-color:#FFF;border-color:inherit;color:#00E;font-weight:bold;text-align:center;vertical-align:top}
+</style> -->
+<table class="tg">
+<thead>
+  <tr>
+    <th class="tg-fr9f"></th>
+    <th class="tg-fr9f"><span style="font-style:normal">Original (darknet)</span></th>
+    <th class="tg-y5w1">tztztztztz/yolov2.pytorch</th>
+    <th class="tg-y5w1">zjykzj/YOLOv2(This)</th>
+    <th class="tg-y5w1">zjykzj/YOLOv2(This)</th>
+    <th class="tg-c3ow">zjykzj/YOLOv2(This)</th>
+  </tr>
+</thead>
+<tbody>
+  <tr>
+    <td class="tg-fr9f">ARCH</td>
+    <td class="tg-zkss">YOLOv2</td>
+    <td class="tg-zkss">YOLOv2</td>
+    <td class="tg-zkss">YOLOv2+Darknet53</td>
+    <td class="tg-zkss">YOLOv2</td>
+    <td class="tg-zkss">YOLOv2-tiny</td>
+  </tr>
+  <tr>
+    <td class="tg-fr9f">VOC AP[IoU=0.50]</td>
+    <td class="tg-zkss">76.8</td>
+    <td class="tg-zkss">72.7</td>
+    <td class="tg-zkss">74.95/76.33(<a href="https://github.com/zjykzj/YOLOv2/releases/tag/v0.2.1">v0.2.1</a>)</td>
+    <td class="tg-zkss">73.27</td>
+    <td class="tg-c3ow">65.44</td>
+  </tr>
+</tbody>
+</table>
+
+* 使用`COCO train2017`数据集进行训练，使用`COCO val2017`数据集进行测试，输入大小为`416x416`。测试结果如下：（*注意：原始论文使用`COCO test-dev2015`的评估结果*）
+
+<!-- <style type="text/css">
+.tg  {border-collapse:collapse;border-spacing:0;}
+.tg td{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
+  overflow:hidden;padding:10px 5px;word-break:normal;}
+.tg th{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
+  font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;}
+.tg .tg-zkss{background-color:#FFF;border-color:inherit;color:#333;text-align:center;vertical-align:top}
+.tg .tg-c3ow{border-color:inherit;text-align:center;vertical-align:top}
+.tg .tg-fr9f{background-color:#FFF;border-color:inherit;color:#333;font-weight:bold;text-align:center;vertical-align:top}
+.tg .tg-y5w1{background-color:#FFF;border-color:inherit;color:#00E;font-weight:bold;text-align:center;vertical-align:top}
+.tg .tg-9y4h{background-color:#FFF;border-color:inherit;color:#1F2328;text-align:center;vertical-align:middle}
+</style> -->
+<table class="tg">
+<thead>
+  <tr>
+    <th class="tg-fr9f"></th>
+    <th class="tg-fr9f"><span style="font-style:normal">Original (darknet)</span></th>
+    <th class="tg-y5w1">zjykzj/YOLOv2(This)</th>
+    <th class="tg-y5w1">zjykzj/YOLOv2(This)</th>
+  </tr>
+</thead>
+<tbody>
+  <tr>
+    <td class="tg-fr9f">ARCH</td>
+    <td class="tg-zkss">YOLOv2</td>
+    <td class="tg-zkss">YOLOv2+Darknet53</td>
+    <td class="tg-zkss">YOLOv2</td>
+  </tr>
+  <tr>
+    <td class="tg-fr9f">COCO AP[IoU=0.50:0.95]</td>
+    <td class="tg-zkss">21.6</td>
+    <td class="tg-9y4h">25.86</td>
+    <td class="tg-9y4h">22.84</td>
+  </tr>
+  <tr>
+    <td class="tg-fr9f">COCO AP[IoU=0.50]</td>
+    <td class="tg-c3ow">44.0</td>
+    <td class="tg-9y4h">48.40</td>
+    <td class="tg-9y4h">43.95</td>
+  </tr>
+</tbody>
+</table>
+
+## 内容列表
+
+- [内容列表](#内容列表)
+- [最近新闻](#最近新闻)
+- [背景](#背景)
+- [数据准备](#数据准备)
+  - [Pascal VOC](#pascal-voc)
+- [安装](#安装)
+  - [需求](#需求)
+  - [容器](#容器)
+- [用法](#用法)
+  - [训练](#训练)
+  - [评估](#评估)
+  - [示例](#示例)
+- [主要维护人员](#主要维护人员)
+- [致谢](#致谢)
+- [参与贡献方式](#参与贡献方式)
+- [许可证](#许可证)
+
+## 最近新闻
+
+* ***[2023/06/28][v0.2.1](https://github.com/zjykzj/YOLOv2/releases/tag/v0.2.1). 重构数据模块。***
+* ***[2023/05/21][v0.2.0](https://github.com/zjykzj/YOLOv2/releases/tag/v0.2.0). 重构损失函数，并且增加了Darknet-53作为Backbone。***
+* ***[2023/05/09][v0.1.2](https://github.com/zjykzj/YOLOv2/releases/tag/v0.1.2). 更新COCO数据集和VOC数据集的训练结果。***
+* ***[2023/05/03][v0.1.1](https://github.com/zjykzj/YOLOv2/releases/tag/v0.1.1). 修复转换函数，并且更新了`yolov2_voc.cfg`和`yolov2-tiny_voc.cfg`在VOC2007 Test上的训练结果。***
+* ***[2023/05/02][v0.1.0](https://github.com/zjykzj/YOLOv2/releases/tag/v0.1.0). 完成了YOLOv2的训练/评估/预测功能，同时提供了VOC2007 Test的测试结果。***
+
+## 背景
+
+YOLOv2在YOLOv1的基础上增加了很多的创新。对于网络结构，它创建了Darknet-19；对于损失函数，他设置锚点框来帮助网络更好的训练。相比于YOLOv1，YOLOv2拥有更好的性能。
+
+本仓库参考了很多的实现，包括[tztztztztz/yolov2.pytorch](https://github.com/tztztztztz/yolov2.pytorch)、[yjh0410/yolov2-yolov3_PyTorch](https://github.com/yjh0410/yolov2-yolov3_PyTorch)和[zjykzj/YOLOv3](https://github.com/zjykzj/YOLOv3)。
+
+## 数据准备
+
+### Pascal VOC
+
+使用脚本[voc2yolov5.py](https://github.com/zjykzj/vocdev/blob/master/py/voc2yolov5.py)
+
+```shell
+python voc2yolov5.py -s /home/zj/data/voc -d /home/zj/data/voc/voc2yolov5-train -l trainval-2007 trainval-2012
+python voc2yolov5.py -s /home/zj/data/voc -d /home/zj/data/voc/voc2yolov5-val -l test-2007
+```
+
+然后将数据集所在的文件夹软链接到指定位置：
+
+```shell
+ln -s /path/to/voc /path/to/YOLOv1/../datasets/voc
+```
+
+## 安装
+
+### 需求
+
+查看[NVIDIA/apex](https://github.com/NVIDIA/apex)
+
+### 容器
+
+开发环境（使用nvidia docker容器）
+
+```shell
+docker run --gpus all -it --rm -v </path/to/YOLOv1>:/app/YOLOv1 -v </path/to/voc>:/app/datasets/voc nvcr.io/nvidia/pytorch:22.08-py3
+```
+
+## 用法
+
+### 训练
+
+* 单个GPU
+
+```shell
+CUDA_VISIBLE_DEVICES=0 python main_amp.py -c configs/yolov1_s14_voc.cfg --opt-level=O1 ../datasets/voc
+```
+
+* 多个GPUs
+
+```shell
+CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch --nproc_per_node=4 --master_port "32111" main_amp.py -c configs/yolov2_voc.cfg --opt-level=O1 ../datasets/voc
+```
+
+### 评估
+
+```shell
+python eval.py -c configs/yolov2_d53_voc.cfg -ckpt outputs/yolov2_d53_voc/model_best.pth.tar ../datasets/voc
+VOC07 metric? Yes                                                                                                                                                                                                  
+AP for aeroplane = 0.7804                                                                                                                                                                                          
+AP for bicycle = 0.8453                                                                                                                                                                                            
+AP for bird = 0.7612                                                                                                                                                                                               
+AP for boat = 0.6260                                                                                                                                                                                               
+AP for bottle = 0.5240                                                                                                                                                                                             
+AP for bus = 0.8261                                                                                                                                                                                                
+AP for car = 0.8244                                                                                                                                                                                                
+AP for cat = 0.8635
+AP for chair = 0.5690
+AP for cow = 0.8161
+AP for diningtable = 0.7046
+AP for dog = 0.8470
+AP for horse = 0.8398
+AP for motorbike = 0.8014
+AP for person = 0.7673
+AP for pottedplant = 0.5069
+AP for sheep = 0.7639
+AP for sofa = 0.7374
+AP for train = 0.8268
+AP for tvmonitor = 0.7581
+Mean AP = 0.7495
+python eval.py -c configs/yolov2_voc.cfg -ckpt outputs/yolov2_voc/model_best.pth.tar ../datasets/voc
+VOC07 metric? Yes
+AP for aeroplane = 0.7396
+AP for bicycle = 0.7876
+AP for bird = 0.7264
+AP for boat = 0.6345
+AP for bottle = 0.4606
+AP for bus = 0.7885
+AP for car = 0.7927
+AP for cat = 0.8630
+AP for chair = 0.5502
+AP for cow = 0.8029
+AP for diningtable = 0.7024
+AP for dog = 0.8457
+AP for horse = 0.8374
+AP for motorbike = 0.8048
+AP for person = 0.7514
+AP for pottedplant = 0.4933
+AP for sheep = 0.7716
+AP for sofa = 0.7068
+AP for train = 0.8618
+AP for tvmonitor = 0.7328
+Mean AP = 0.7327
+python eval.py -c configs/yolov2-tiny_voc.cfg -ckpt outputs/yolov2-tiny_voc/model_best.pth.tar ../datasets/voc
+VOC07 metric? Yes
+AP for aeroplane = 0.6745
+AP for bicycle = 0.7511
+AP for bird = 0.6245
+AP for boat = 0.5421
+AP for bottle = 0.3319
+AP for bus = 0.7508
+AP for car = 0.7413
+AP for cat = 0.8123
+AP for chair = 0.4276
+AP for cow = 0.7286
+AP for diningtable = 0.6336
+AP for dog = 0.7646
+AP for horse = 0.8083
+AP for motorbike = 0.7378
+AP for person = 0.6835
+AP for pottedplant = 0.3593
+AP for sheep = 0.6390
+AP for sofa = 0.6519
+AP for train = 0.7772
+AP for tvmonitor = 0.6479
+Mean AP = 0.6544
+python eval.py -c configs/yolov2_d53_coco.cfg -ckpt outputs/yolov2_d53_coco/model_best.pth.tar --traversal ../datasets/coco
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.259
+ Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.484
+ Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.249
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.071
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.285
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.440
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.233
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.341
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.351
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.110
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.387
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.585
+python eval.py -c configs/yolov2_coco.cfg -ckpt outputs/yolov2_coco/model_best.pth.tar --traversal ../datasets/coco
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.228
+ Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.440
+ Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.217
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.052
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.237
+ Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.402
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.216
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.313
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.324
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.092
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.344
+ Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.561
+```
+
+### 示例
+
+```shell
+python demo.py -c 0.6 configs/yolov2_voc.cfg outputs/yolov2_voc/model_best.pth.tar --exp voc assets/voc2007-test/
+```
+
+<p align="left"><img src="results/voc/000237.jpg" height="240"\>  <img src="results/voc/000386.jpg" height="240"\></p>
+
+```shell
+python demo.py -c 0.6 configs/yolov2_coco.cfg outputs/yolov2_coco/model_best.pth.tar --exp coco assets/coco/
+```
+
+<p align="left"><img src="results/coco/bus.jpg" height="240"\>  <img src="results/coco/zidane.jpg" height="240"\></p>
+
+## 主要维护人员
+
+* zhujian - *Initial work* - [zjykzj](https://github.com/zjykzj)
+
+## 致谢
+
+* [tztztztztz/yolov2.pytorch](https://github.com/tztztztztz/yolov2.pytorch)
+* [AlexeyAB/darknet](https://github.com/AlexeyAB/darknet)
+* [yjh0410/yolov2-yolov3_PyTorch](https://github.com/yjh0410/yolov2-yolov3_PyTorch)
+* [zjykzj/YOLOv3](https://github.com/zjykzj/YOLOv3)
+* [zjykzj/anchor-boxes](https://github.com/zjykzj/anchor-boxes)
+* [zjykzj/vocdev](https://github.com/zjykzj/vocdev)
+* [ultralytics/yolov5](https://github.com/ultralytics/yolov5)
+
+## 参与贡献方式
+
+欢迎任何人的参与！打开[issue](https://github.com/zjykzj/YOLOv2/issues)或提交合并请求。
+
+注意:
+
+* `GIT`提交，请遵守[Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0-beta.4/)规范
+* 语义版本化，请遵守[Semantic Versioning 2.0.0](https://semver.org)规范
+* `README`编写，请遵守[standard-readme](https://github.com/RichardLitt/standard-readme)规范
+
+## 许可证
+
+[Apache License 2.0](LICENSE) © 2023 zjykzj