README.md 3.64 KB
Newer Older
sunxx1's avatar
sunxx1 committed
1
2
# Mobilenetv2

sunxx1's avatar
sunxx1 committed
3
4
5
6
7
8
## 论文

MobileNetV2: Inverted Residuals and Linear Bottlenecks

- https://openaccess.thecvf.com/content_cvpr_2018/papers/Sandler_MobileNetV2_Inverted_Residuals_CVPR_2018_paper.pdf

sunxx1's avatar
sunxx1 committed
9
## 模型结构
sunxx1's avatar
sunxx1 committed
10
11
12

MobileNetV2是一种轻量级的卷积神经网络模型,由Google在2018年提出。它是MobileNet系列中的第二个版本,主要用于移动设备和嵌入式设备等资源受限的环境中进行图像分类、目标检测等计算机视觉任务。

renzhc's avatar
renzhc committed
13
![loading-ag-535](./images/d15a0e56517b4f7284a862f1d6eaef9a.png)
sunxx1's avatar
sunxx1 committed
14
15
16



sunxx1's avatar
sunxx1 committed
17
## 算法原理
sunxx1's avatar
sunxx1 committed
18

sunxx1's avatar
sunxx1 committed
19
MobileNetV2的网络结构主要由两部分组成:特征提取层和分类器。
sunxx1's avatar
sunxx1 committed
20

sunxx1's avatar
sunxx1 committed
21
![20231124104337](./images/20231124104337.png)
sunxx1's avatar
sunxx1 committed
22

sunxx1's avatar
sunxx1 committed
23
## 环境配置
sunxx1's avatar
sunxx1 committed
24

sunxx1's avatar
sunxx1 committed
25
### Docker(方法一)
sunxx1's avatar
sunxx1 committed
26

sunxx1's avatar
sunxx1 committed
27
28
```python
git clone --recursive http://developer.hpccube.com/codes/modelzoo/mobilenetv2_mmcv.git
renzhc's avatar
renzhc committed
29
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.1.0-ubuntu20.04-dtk24.04.1-py3.10
sunxx1's avatar
sunxx1 committed
30
# <your IMAGE ID>用以上拉取的docker的镜像ID替换
renzhc's avatar
renzhc committed
31
docker run -it --shm-size 16g --network=host --name=mobilenetv2 --privileged --device=/dev/kfd --device=/dev/dri --device=/dev/mkfd --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v /opt/hyhal:/opt/hyhal:ro -v $PWD/mobilenetv2_mmcv:/home/mobilenetv2_mmcv <your IMAGE ID> bash
sunxx1's avatar
sunxx1 committed
32

renzhc's avatar
renzhc committed
33
cd mobilenetv2_mmcv/mmpretrain-mmcv
sunxx1's avatar
sunxx1 committed
34
35
36
37
38
39
40
41
pip install -r requirements.txt
```

### Dockerfile(方法二)

```plaintext
cd mobilenetv2_mmcv/docker
docker build --no-cache -t mobilenetv2_mmcv:latest .
renzhc's avatar
renzhc committed
42
docker run -it --shm-size 16g --network=host --name=mobilenetv2 --privileged --device=/dev/kfd --device=/dev/dri --device=/dev/mkfd --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v /opt/hyhal:/opt/hyhal:ro -v $PWD/mobilenetv2_mmcv:/home/mobilenetv2_mmcv <your IMAGE ID> bash
sunxx1's avatar
sunxx1 committed
43
44
45
46
47
48
49
50
# 若遇到Dockerfile启动的方式安装环境需要长时间等待,可注释掉里面的pip安装,启动容器后再安装python库:pip install -r requirements.txt
```

### Anaconda(方法三)

1、关于本项目DCU显卡所需的特殊深度学习库可从光合开发者社区下载安装: https://developer.hpccube.com/tool/

```plaintext
renzhc's avatar
renzhc committed
51
52
53
54
55
56
DTK驱动: dtk24.04.1
python: python3.10
torch: 2.1.0
torchvision: 0.16.0
mmcv: 2.0.1
mmengine: 0.10.4
sunxx1's avatar
sunxx1 committed
57
58
59
60
Tips:以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应
```

2、其它非特殊库参照requirements.txt安装
sunxx1's avatar
sunxx1 committed
61

sunxx1's avatar
sunxx1 committed
62
```plaintext
sunxx1's avatar
sunxx1 committed
63
64
pip install -r requirements.txt
```
sunxx1's avatar
sunxx1 committed
65

sunxx1's avatar
sunxx1 committed
66
## 数据集
sunxx1's avatar
sunxx1 committed
67

sunxx1's avatar
sunxx1 committed
68
69
在本测试中可以使用ImageNet数据集。

sunxx1's avatar
sunxx1 committed
70
71
72
73
下载ImageNet数据集:https://image-net.org/

下载val数据:链接:https://pan.baidu.com/s/1oXsmsYahGVG3uOZ8e535LA?pwd=c3bc 提取码:c3bc 替换ImageNet数据集中的val目录,处理后的数据结构如下:

sunxx1's avatar
sunxx1 committed
74
```
dcuai's avatar
dcuai committed
75
76
77
78
79
80
81
82
data
    ├──imagenet
        ├── meta
            ├──val.txt
            ├──train.txt
            ...
        ├── train
        ├── val
renzhc's avatar
renzhc committed
83

sunxx1's avatar
sunxx1 committed
84
```
renzhc's avatar
renzhc committed
85

dongchy920's avatar
dongchy920 committed
86
87
88
SCNet快速下载链接[http://113.200.138.88:18080/aidatasets/project-dependency/imagenet-2012
](http://113.200.138.88:18080/aidatasets/project-dependency/imagenet-2012
)
renzhc's avatar
renzhc committed
89
90
91

如果imagenet数据集较大,则可以使用tiny-imagenet-200,但此时需要对配置文件进行修改。

sunxx1's avatar
sunxx1 committed
92
## 训练
sunxx1's avatar
sunxx1 committed
93
94
95

将训练数据解压到data目录下。

sunxx1's avatar
sunxx1 committed
96
### 单机8卡
sunxx1's avatar
sunxx1 committed
97

renzhc's avatar
renzhc committed
98
    bash tools/dist_train.sh <PYTHON配置文件> 8 
sunxx1's avatar
sunxx1 committed
99

dcuai's avatar
dcuai committed
100
101
102
103
104
## result

![img](https://developer.hpccube.com/codes/modelzoo/vit_pytorch/-/raw/master/image/README/1695381570003.png)

### 精度
sunxx1's avatar
sunxx1 committed
105

renzhc's avatar
renzhc committed
106
未测试
sunxx1's avatar
sunxx1 committed
107

sunxx1's avatar
sunxx1 committed
108
109
110
111
112
113
114
115
## 应用场景

### 算法类别

图像分类

### 热点行业

sunxx1's avatar
sunxx1 committed
116
制造,能源,交通,网安
sunxx1's avatar
sunxx1 committed
117

dcuai's avatar
dcuai committed
118
## 源码仓库及问题反馈
sunxx1's avatar
sunxx1 committed
119

sunxx1's avatar
sunxx1 committed
120
https://developer.hpccube.com/codes/modelzoo/mobilenetv2_mmcv
sunxx1's avatar
sunxx1 committed
121

dcuai's avatar
dcuai committed
122
## 参考资料
sunxx1's avatar
sunxx1 committed
123

dcuai's avatar
dcuai committed
124
https://github.com/open-mmlab/mmpretrain