README.md 5.53 KB
Newer Older
dongchy920's avatar
dongchy920 committed
1
2
3
4
5
6
7
8
9
10
11
12
# ArcFace
## 论文
- https://arxiv.org/pdf/1801.07698.pdf
## 模型结构
这篇文章提出一种新的用于人脸识别的损失函数:additive angular margin loss,直接在角度空间(angular space)中最大化分类界限,基于该损失函数训练得到人脸识别算法ArcFace。
<div align=center>
    <img src="./docs/arcface.png"/>
</div>

## 算法原理
通过训练深度卷积神经网络嵌入 (DCNN Embedding) 来进行人脸识别。  
ArcFace训练流程:  
dongchy920's avatar
dongchy920 committed
13
<div align=center>
dongchy920's avatar
dongchy920 committed
14
<<<<<<< HEAD
dongchy920's avatar
dongchy920 committed
15
    <img src="yuanli.png"/>
dongchy920's avatar
dongchy920 committed
16
17
18
19
20
21
22
=======
<<<<<<< HEAD
    <img src="yuanli.png"/>
=======
    <img src="./docs/legend.png"/>
>>>>>>> e9e2958 (arcface_pytorch_v2)
>>>>>>> 0ed37c0 (readme modify)
dongchy920's avatar
dongchy920 committed
23
</div>  
dongchy920's avatar
dongchy920 committed
24
25
26
27
28
29
30
31
32

<div align=center>
    <img src="./docs/train.jpg"/>
</div>


## 环境配置
### Docker(方法一)
[光源](https://www.sourcefind.cn/#/service-list)中拉取docker镜像:
dongchy920's avatar
arcface  
dongchy920 committed
33
```
dongchy920's avatar
dongchy920 committed
34
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:1.13.1-centos7.6-dtk23.10-py310
dongchy920's avatar
arcface  
dongchy920 committed
35
```
dongchy920's avatar
dongchy920 committed
36
创建容器并挂载目录进行开发:
dongchy920's avatar
arcface  
dongchy920 committed
37
```
dongchy920's avatar
dongchy920 committed
38
39
40
41
42
docker run -it --name {name} --shm-size=1024G  --device=/dev/kfd --device=/dev/dri/ --privileged --cap-add=SYS_PTRACE --security-opt seccomp=unconfined --ulimit memlock=-1:-1 --ipc=host --network host --group-add video -v /opt/hyhal:/opt/hyhal:ro -v {}:{} {docker_image} /bin/bash
# 修改1 {name} 需要改为自定义名称,建议命名{框架_dtk版本_使用者姓名},如果有特殊用途可在命名框架前添加命名
# 修改2 {docker_image} 需要需要创建容器的对应镜像名称,如: pytorch:1.10.0-centos7.6-dtk-23.04-py37-latest【镜像名称:tag名称】
# 修改3 -v 挂载路径到容器指定路径
pip install -r requirements.txt
dongchy920's avatar
arcface  
dongchy920 committed
43
```
dongchy920's avatar
dongchy920 committed
44
45
46
47
48
49
50
51
52
53
54
55
56
### Dockerfile(方法二)
```
cd docker
docker build --no-cache -t arcface_pytorch:1.0 .
docker run -it --name {name} --shm-size=1024G  --device=/dev/kfd --device=/dev/dri/ --privileged --cap-add=SYS_PTRACE --security-opt seccomp=unconfined --ulimit memlock=-1:-1 --ipc=host --network host --group-add video -v /opt/hyhal:/opt/hyhal:ro -v {}:{} {docker_image} /bin/bash 
pip install -r requirements.txt
```
### Anaconda(方法三)
线上节点推荐使用conda进行环境配置。
创建python=3.10的conda环境并激活
```
conda create -n arcface python=3.10
conda activate arcface
dongchy920's avatar
arcface  
dongchy920 committed
57
58
```

dongchy920's avatar
dongchy920 committed
59
60
61
62
63
64
65
66
67
68
69
70
71
关于本项目DCU显卡所需的特殊深度学习库可从[光合](https://developer.hpccube.com/tool/)开发者社区下载安装。
```
DTK驱动:dtk23.10
python:python3.10
pytorch:1.13.1
torchvision:0.14.1
```
安装其他依赖包
```
pip install -r requirements.txt
```
## 数据集
`MS1MV2\IJBC`
dongchy920's avatar
arcface  
dongchy920 committed
72

dongchy920's avatar
dongchy920 committed
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
- 训练集[faces_emore.zip](https://pan.baidu.com/s/1S6LJZGdqcZRle1vlcMzHOQ)
下载后解压到当前目录  
数据目录结构如下:
```
 ── faces_emore
    |   agedb_30.bin
    |   calfw.bin
    |   cfp_ff.bin
    |   cfp_fp.bin
    |   cplfw.bin
    |   lfw.bin
    |   property
    |   train.idx
    |   train.rec
    |   vgg2_fp.bin
```
- 测试集[IJBC.zip](https://pan.baidu.com/s/1Ok4sqTO8vqAE_kG3zV1rqw?pwd=1234)  
解压分卷压缩文件:
```
# 将所有的分卷压缩文件放在一个文件夹中
zip -s 0 IJBC.zip --out IJBC_ALL.zip
dongchy920's avatar
dongchy920 committed
94
unzip IJBC_ALL.zip
dongchy920's avatar
dongchy920 committed
95
```
dongchy920's avatar
arcface  
dongchy920 committed
96

dongchy920's avatar
dongchy920 committed
97
98
## 训练
Backbone使用ResNet100,在MS1MV3数据集上的预训练权重文件为[model.pt](https://pan.baidu.com/s/1W-TisIZtZmRQz32hq5T6Uw?pwd=1234)  
dongchy920's avatar
arcface  
dongchy920 committed
99

dongchy920's avatar
dongchy920 committed
100
101
### 单机单卡
```
dongchy920's avatar
dongchy920 committed
102
103
104
105
<<<<<<< HEAD
python train_v2.py configs/ms1mv2_r100
=======
<<<<<<< HEAD
dongchy920's avatar
dongchy920 committed
106
python train_v2.py configs/ms1mv2_r100
dongchy920's avatar
dongchy920 committed
107
108
109
110
=======
python train_v2.py configs/ms1mv2_r100.py
>>>>>>> e9e2958 (arcface_pytorch_v2)
>>>>>>> 0ed37c0 (readme modify)
dongchy920's avatar
dongchy920 committed
111
```
dongchy920's avatar
arcface  
dongchy920 committed
112

dongchy920's avatar
dongchy920 committed
113
114
### 单机多卡
```
dongchy920's avatar
dongchy920 committed
115
<<<<<<< HEAD
dongchy920's avatar
dongchy920 committed
116
117
torchrun --nproc_per_node=4 train_v2.py configs/ms1mv2_r100
```
dcuai's avatar
dcuai committed
118
### 测试
dongchy920's avatar
dongchy920 committed
119
120
121
122
123
124
125
126
127
128
129
=======
<<<<<<< HEAD
torchrun --nproc_per_node=4 train_v2.py configs/ms1mv2_r100
```
### 测试
=======
torchrun --nproc_per_node=4 train_v2.py configs/ms1mv2_r100.py
```
## 测试
>>>>>>> e9e2958 (arcface_pytorch_v2)
>>>>>>> 0ed37c0 (readme modify)
dongchy920's avatar
dongchy920 committed
130
131
132
133
下载权重文件和测试数据集,测试模型精度:
```
python eval_ijbc.py --model-prefix model.pt --image-path IJBC_ALL --network r100
```
dongchy920's avatar
dongchy920 committed
134
135
136
137
<<<<<<< HEAD
=======
<<<<<<< HEAD
>>>>>>> 0ed37c0 (readme modify)
dcuai's avatar
dcuai committed
138
139
140
141
## result

### 精度

dongchy920's avatar
dongchy920 committed
142
143
144
145
146
147
148
模型在MS1MV2数据集的测试指标:
| 模型 | 数据类型 | AUC |
| :------: | :------: | :------: |
| [r34](https://pan.baidu.com/s/1LR0zm8AxwN2tZH55xQdzHw?pwd=1234) | fp16 | 99.4611% |
| [r50](https://pan.baidu.com/s/128GP5J-jWvNbQAAur68bHw?pwd=1234) | fp16 | 99.4854% |
| [r100](https://pan.baidu.com/s/1cslUcKgv5dSrJtBp62J6Fw?pwd=1234) | fp16 | 99.5296% |
| [r100](https://pan.baidu.com/s/1KRBAKFzJU2ZOqhVHe91N6A?pwd=1234) | fp32 | 99.5612% |
dongchy920's avatar
dongchy920 committed
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
<<<<<<< HEAD
=======
=======

## result
<div align=center>
    <img src="./docs/ROC.png"/>
</div>

## 精度
模型在MS1MV2数据集的测试指标:
| 模型 | 数据类型 | AUC |
| :------: | :------: | :------: |
| [r34](https://pan.baidu.com/s/1LR0zm8AxwN2tZH55xQdzHw?pwd=1234) | fp16 | 99.5611% |
| [r50](https://pan.baidu.com/s/128GP5J-jWvNbQAAur68bHw?pwd=1234) | fp16 | 99.5616% |
| [r100](https://pan.baidu.com/s/1e4Qg2i6wqyBCcwgA-dA8xw?pwd=1234) | fp16 | 99.5767% |
>>>>>>> e9e2958 (arcface_pytorch_v2)
>>>>>>> 0ed37c0 (readme modify)
dongchy920's avatar
arcface  
dongchy920 committed
167
168


dongchy920's avatar
dongchy920 committed
169
170
171
## 应用场景
### 算法类别
人脸识别
dongchy920's avatar
arcface  
dongchy920 committed
172

dongchy920's avatar
dongchy920 committed
173
174
### 热点应用行业
安防,交通,教育
dongchy920's avatar
arcface  
dongchy920 committed
175
176


dongchy920's avatar
dongchy920 committed
177
178
179
180
## 源码仓库及问题反馈
[https://developer.hpccube.com/codes/modelzoo/arcface_pytorch](https://developer.hpccube.com/codes/modelzoo/arcface_pytorch)
## 参考资料
[https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch](https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch)
dongchy920's avatar
arcface  
dongchy920 committed
181
182