• Deep-Learning-Models-For-Classification
• • Introduction
  • configures

Deep-Learning-Models-For-Classification

This readme is introduced in Chinese (including most of the comments in the code). Please translate it into English if necessary.

Introduction

以下是本项目支持的模型列表，包含了自AelxNet以来经典的深度学习分类模型，大部分模型是基于卷积神经网络的，也有一部分是基于注意力机制的。

模型代码在classic_models文件夹中。博客链接是对模型的介绍，有一些正在编写，会持续更新...

This project organizes classic classification Neural Networks based convolution or attention mechanism:

1. AlexNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123839843

2. ZFNet
   Blog Introduction Link: WRINTING

3. VggNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123716634

4. GoogleNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123717625

5. ResNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123739792

6. DenseNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123765554

7. MobileNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123793236

8. ShuffleNet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123797686

9. SENet
   Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123848298

10. Vision_Transformer
    Blog Introduction Link: WRINTING

11. Swin_Transformer
    Blog Introduction Link: WRINTING

12. EfficientNet
    Blog Introduction Link: https://blog.csdn.net/qq_39297053/article/details/123804502

13. ConvNeXt
    Blog Introduction Link: WRINTIN

14. MLP-mixer
    Blog Introduction Link: WRINTING... ...

In additon, I write training and inference python scripts for image classification task.
train.py

configures

本项目是使用python语言基于pytorch深度学习框架编写的。
此外，我写了三个训练脚本用于模型的训练，默认的数据集是花朵数据集，此数据集包含五种不同种类共三千多张花朵图像，下载链接：链接：https://pan.baidu.com/s/1EhPMVLOQlLNN55ndrLbh4Q
提取码：7799 。如要使用，请指定参超到数据集地址/flower（eg： --data_path /.../.../.../flower）

三个训练脚本中，train_sample.py是最简单的实现；train.py是升级版的实现，具体改进点见train.py脚本中的注释; train_distrubuted.py支持多gpu分布式训练。

最后，test.py是推理脚本。dataload中是数据集加载代码；utils是封装的功能包，包括学习策略，训练和验证，分布式初始化，可视化等等。