【人脸检测】 Tinaface复现（数据集准备、测试与评估）

ゝ一世哀愁。 2022-12-25 09:56 149阅读 0赞

近期代码貌似作者有改动，想复现可以参考本文环境。具体细节变通一下。

参考：

[TinaFace: Strong but Simple Baseline for Face Detection][TinaFace_ Strong but Simple Baseline for Face Detection]

[https://github.com/Media-Smart/vedadet][https_github.com_Media-Smart_vedadet]

[https://github.com/sovrasov/wider-face-pascal-voc-annotations][https_github.com_sovrasov_wider-face-pascal-voc-annotations]

# 0.环境 #

ubuntu16.04
    python3.6
    cuda9.2
    cudnn7
    
    torch==1.6.0+cu92
    
    addict==2.4.0
    cycler==0.10.0
    Cython==0.29.21
    future==0.18.2
    kiwisolver==1.3.1
    matplotlib==3.3.3
    numpy==1.19.4
    Pillow==8.0.1
    pycurl==7.43.0
    pygobject==3.20.0
    pyparsing==2.4.7
    python-apt==1.1.0b1+ubuntu0.16.4.9
    python-dateutil==2.8.1
    six==1.15.0
    unattended-upgrades==0.1
    
    tqdm
    scipy
    ipython

cd vedadet
    pip install -r requirements/build.txt

避免出错，我直接安装环境中对应的python库，然后注释掉对应的根目录下setup.py中的安装依赖包的两行。

![20201203104255225.png][]

改为：

![20201203104221274.png][]

（默认会安装最新的，如果对应的cuda版本不支持的话，那就自己安装吧！！！）然后执行下面：

pip install -v -e .

# 1.准备 #

## 1.1 准备模型 ##

[https://drive.google.com/u/0/uc?id=1zU738coEVDBkLBUa4hvJUucL7dcSBT7v&export=download][https_drive.google.com_u_0_uc_id_1zU738coEVDBkLBUa4hvJUucL7dcSBT7v_export_download]

可以将放到[https://github.com/Media-Smart/vedadet/tree/main/configs/trainval/tinaface][https_github.com_Media-Smart_vedadet_tree_main_configs_trainval_tinaface]，目录下新建的models下。

## 1.2 准备数据 ##

### 1）原始数据 ###

[https://github.com/Media-Smart/vedadet/tree/main/configs/trainval/tinaface][https_github.com_Media-Smart_vedadet_tree_main_configs_trainval_tinaface]

进入vedadet/目录下，新建data，然后进入data目录下，从widerface官网上下载训练、验证与测试数据：

![20201203110719136.png][]

解压从官网下载的数据，目录应该是这样：

vedadet
    ├── vedadet
    ├── vedacore
    ├── tools
    ├── configs
    ├── data
    │   ├── WIDERFace
    │   │   ├── WIDER_train
    │   │   │   ├── images
    │   │   │   │      ├── 0--Parade
    │   │   │   │      ├── ......
    │   │   │   │      ├── 61--Street_Battle
    │   │   ├── WIDER_val
    │   │   │   ├── images
    │   │   │   │      ├── 0--Parade
    │   │   │   │      ├── ......
    │   │   │   │      ├── 61--Street_Battle
    │   │   ├── WIDER_test
    │   │   ├── wider_face_split

### 2）xml文件 ###

下载对应的VOC格式的xml文件：

git clone https://github.com/sovrasov/wider-face-pascal-voc-annotations.git
    cd vedadet/data
    mv wider-face-pascal-voc-annotations/ WIDERFace/
    mv WIDERFace/WIDER_train_annotations WIDERFace/WIDER_train/Annotations
    mv WIDERFace/WIDER_val_annotations WIDERFace/WIDER_val/Annotations

然后利用命令：

cd ..
    sh configs/trainval/tinaface/gen_xml_name_txt.sh

目录结构如下：

vedadet
    ├── vedadet
    ├── vedacore
    ├── tools
    ├── configs
    ├── data
    │   ├── WIDERFace
    │   │   ├── WIDER_train
    │   │   │   ├── images
    │   │   │   │      ├── 0--Parade
    │   │   │   │      ├── ......
    │   │   │   │      ├── 61--Street_Battle
    │   │   │   ├── Annotations
    │   │   ├── WIDER_val
    │   │   │   ├── images
    │   │   │   │      ├── 0--Parade
    │   │   │   │      ├── ......
    │   │   │   │      ├── 61--Street_Battle
    │   │   │   ├── Annotations
    │   │   ├── WIDER_test
    │   │   ├── wider_face_split

### 3）代码中路径等修改 ###

由于与作者给出的目录略微有点不同，所以修改以下：

# \vedadet\configs\trainval\tinaface\tinaface.py line 14与38
            img_prefix=data_root + 'WIDER_train/',
            img_prefix=data_root + 'WIDER_val/',
    改为：
            img_prefix=data_root + 'WIDER_train/images/',
            img_prefix=data_root + 'WIDER_val/images/',
    
    # \vedadet\vedadet\datasets\widerface.py line 36-37
                xml_path = osp.join(self.img_prefix, 'Annotations',
                                    f'{img_id}.xml')
    改为
                xml_path = osp.join(self.img_prefix.replace('images','Annotations'),
                                    f'{img_id}.xml')

！！！其中 \\vedadet\\vedadet\\datasets\\xml\_style.py 中对应的xml\_path也要修改为相应的。

至此，基本就可以了。如果想修改数据的根目录：

# \vedadet\configs\trainval\tinaface\tinaface.py line 3 改为你对应的对应就可以了，绝对与相对路径都可以
    data_root = './data/WIDERFace/'

# 2.测试与评估 #

## 2.1 测试 ##

python configs/trainval/tinaface/test_widerface.py configs/trainval/tinaface/tinaface.py /root/vedadet/weights/tinaface_r50_fpn_widerface.pth

会在vedadet/eval\_dirs/tmp/tinaface/目录下，生成对应的txt文件。

## 2.2 评估 ##

还是用之前很多篇评估widerface的方式来评估一下。

将Pytorch\_Retinaface/widerface\_evaluate内的这几个文件与widerface\_eval文件夹，复制到vedadet/tools目录下。

[https://github.com/biubug6/Pytorch\_Retinaface/tree/master/widerface\_evaluate][https_github.com_biubug6_Pytorch_Retinaface_tree_master_widerface_evaluate]

[https://github.com/ChiCheng123/SRN/tree/master/tools][https_github.com_ChiCheng123_SRN_tree_master_tools]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 1][]

![20201201101647943.png][]

![20201203144028858.png][]

widerface\_eval文件也可以在我的资源中下载：[https://download.csdn.net/download/qq\_35975447/14158058][https_download.csdn.net_download_qq_35975447_14158058]

命令加修改：

cd ./tools
    python setup.py build_ext --inplace
     
    vim evaluation.py
    # line 287-288
    parser.add_argument('-p', '--pred', default="${vedadet_root}/eval_dirs/tmp/tinaface/")
    parser.add_argument('-g', '--gt', default='./widerface_eval/ground_truth/')
     
    python evaluation.py

其中参数：

\--pred是预测结果txt文件的保存路径，默认为：

![20210114165606247.png][]

--gt是真实的，我们之前准备的。

## 2.3 评估结果  ##

![2020120315085544.png][]

评估结果还是比较准的，这个是没有加TTA的。

#  参考 #

[1.vedadet（官方代码）][https_github.com_Media-Smart_vedadet]

[2.tinaface（官方论文）][2.tinaface]

[3.wider-face-pascal-voc-annotations（xml文件）][3.wider-face-pascal-voc-annotations_xml]

[4.评估部分参考（SRN）][4._SRN]

[5.评估部分代码与gt][https_github.com_biubug6_Pytorch_Retinaface_tree_master_widerface_evaluate]

[TinaFace_ Strong but Simple Baseline for Face Detection]: https://arxiv.org/pdf/2011.13183.pdf
[https_github.com_Media-Smart_vedadet]: https://github.com/Media-Smart/vedadet
[https_github.com_sovrasov_wider-face-pascal-voc-annotations]: https://github.com/sovrasov/wider-face-pascal-voc-annotations
[20201203104255225.png]: /images/20221120/1f4748d291ad43a29bab1bba19758f7d.png
[20201203104221274.png]: https://img-blog.csdnimg.cn/20201203104221274.png
[https_drive.google.com_u_0_uc_id_1zU738coEVDBkLBUa4hvJUucL7dcSBT7v_export_download]: https://drive.google.com/u/0/uc?id=1zU738coEVDBkLBUa4hvJUucL7dcSBT7v&export=download
[https_github.com_Media-Smart_vedadet_tree_main_configs_trainval_tinaface]: https://github.com/Media-Smart/vedadet/tree/main/configs/trainval/tinaface
[20201203110719136.png]: https://img-blog.csdnimg.cn/20201203110719136.png
[https_github.com_biubug6_Pytorch_Retinaface_tree_master_widerface_evaluate]: https://github.com/biubug6/Pytorch_Retinaface/tree/master/widerface_evaluate
[https_github.com_ChiCheng123_SRN_tree_master_tools]: https://github.com/ChiCheng123/SRN/tree/master/tools
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20210114144559319.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/20210114151326149.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3,size_16,color_FFFFFF,t_70
[20201201101647943.png]: https://img-blog.csdnimg.cn/20201201101647943.png
[20201203144028858.png]: https://img-blog.csdnimg.cn/20201203144028858.png
[https_download.csdn.net_download_qq_35975447_14158058]: https://download.csdn.net/download/qq_35975447/14158058
[20210114165606247.png]: https://img-blog.csdnimg.cn/20210114165606247.png
[2020120315085544.png]: https://img-blog.csdnimg.cn/2020120315085544.png
[2.tinaface]: https://arxiv.org/abs/2011.13183
[3.wider-face-pascal-voc-annotations_xml]: https://github.com/akofman/wider-face-pascal-voc-annotations
[4._SRN]: https://blog.csdn.net/qq_35975447/article/details/110288767