README.md 2.48 KB
Newer Older
sunxx1's avatar
sunxx1 committed
1
2
# Densenet121

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

Densely Connected Convolutional Networks

- https://arxiv.org/pdf/1608.06993.pdf

sunxx1's avatar
sunxx1 committed
9
10
## 模型介绍

sunxx1's avatar
sunxx1 committed
11
12
DenseNet-121是一种深度卷积神经网络,如图所示,由Kaiming He等人于2017年提出。它是DenseNet系列中的一种,也是其中最流行的一种,被广泛应用于计算机视觉领域的图像分类、目标检测和语义分割等任务。

sunxx1's avatar
sunxx1 committed
13
![image-20231120204030674](./images/image-20231120204030674.png)
sunxx1's avatar
sunxx1 committed
14
15
16

## 模型结构

sunxx1's avatar
sunxx1 committed
17
DenseNet的核心组件为“Dense Block”,如图所示,由Dense connectivity和Transition Layer组成。每个密集块中包含若干个卷积层和池化层,每个卷积层都会接收前面所有层的输入,并将它们连接到自己的输出上。而过渡层则用于将前面密集块的输出进行降维,减少参数数量。
sunxx1's avatar
sunxx1 committed
18

sunxx1's avatar
sunxx1 committed
19
![image-20231120204212494](./images/image-20231120204212494.png)
sunxx1's avatar
sunxx1 committed
20

sunxx1's avatar
sunxx1 committed
21
## 环境配置
sunxx1's avatar
sunxx1 committed
22

sunxx1's avatar
sunxx1 committed
23
### Docker(方法一)
sunxx1's avatar
sunxx1 committed
24

sunxx1's avatar
sunxx1 committed
25
```python
sunxx1's avatar
sunxx1 committed
26
git clone --recursive http://developer.hpccube.com/codes/modelzoo/densenet121_mmcv.git
sunxx1's avatar
sunxx1 committed
27
28
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:1.10.0-centos7.6-dtk-22.10.1-py37-latest
# <your IMAGE ID>用以上拉取的docker的镜像ID替换
sunxx1's avatar
sunxx1 committed
29
docker run --shm-size 10g --network=host --name=nit-pytorch --privileged --device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v $PWD/densenet121_mmcv:/home/densenet121_mmcv -it <your IMAGE ID> bash
sunxx1's avatar
sunxx1 committed
30

sunxx1's avatar
sunxx1 committed
31
cd densenet121_mmcv/mmclassification-mmcv
sunxx1's avatar
sunxx1 committed
32
33
pip install -r requirements.txt
```
sunxx1's avatar
sunxx1 committed
34

sunxx1's avatar
sunxx1 committed
35
## 数据集
sunxx1's avatar
sunxx1 committed
36

sunxx1's avatar
sunxx1 committed
37
在本测试中可以使用ImageNet数据集。
sunxx1's avatar
sunxx1 committed
38

sunxx1's avatar
sunxx1 committed
39
40
41
42
43
44
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
45
46
47
48
49
```
├── meta
├── train
├── val
```
sunxx1's avatar
sunxx1 committed
50

sunxx1's avatar
sunxx1 committed
51
### 训练
sunxx1's avatar
sunxx1 committed
52
53
54

将训练数据解压到data目录下。

sunxx1's avatar
sunxx1 committed
55
### 单机8卡
sunxx1's avatar
sunxx1 committed
56
57
58

    ./densenet121.sh

sunxx1's avatar
sunxx1 committed
59
## 精度
sunxx1's avatar
sunxx1 committed
60
61
62

测试数据使用的是ImageNet数据集,使用的加速卡是DCU Z100L。

sunxx1's avatar
sunxx1 committed
63
64
65
66
| 卡数 |           精度            |
| :--: | :-----------------------: |
|  8   | top1:0.74044;top5:0.91672 |

sunxx1's avatar
sunxx1 committed
67
68
69
70
71
72
73
74
75
76
77
78
## result

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

## 应用场景

### 算法类别

图像分类

### 热点行业

sunxx1's avatar
sunxx1 committed
79
制造,能源,交通,网安
sunxx1's avatar
sunxx1 committed
80

sunxx1's avatar
sunxx1 committed
81
### 源码仓库及问题反馈
sunxx1's avatar
sunxx1 committed
82

sunxx1's avatar
sunxx1 committed
83
http://developer.hpccube.com/codes/modelzoo/densenet121_mmcv.git
sunxx1's avatar
sunxx1 committed
84

sunxx1's avatar
sunxx1 committed
85
### 参考
sunxx1's avatar
sunxx1 committed
86

sunxx1's avatar
sunxx1 committed
87
https://github.com/open-mmlab/mmpretrain