CNN Implementation for Fashion-MNIST in PyTorch

Introduction

This repository is the implementation of some classical Convolutional Neural Networks for Fashion-MNIST in PyTorch. Please feel free to ask anything!

• LeNet, Paper Link
• AlexNet, Paper Link
• VGGNet, Paper Link
• InceptionNet, Paper Link
• ResNet, Paper Link

Baseline Environment and Setup

python 3.9.11
cuda 11.0.3
cudnn 8.0.5 
torch 1.7.1

• Install pytorch: pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
• Install dependencies: pip install -r requirements.txt
• Download the Fashion-MNIST then put the unzipped file into the data folder
• Download the best.pth.tar from corresponding pre-trained model parameter folder (seeing the table in Working Record) and put the best.pth.tar file into the model folder
• Choose the model you need in Line125-129 of train.py and Check the hyper parameters

Project Structure

.
|-- config/             
|-- data/
|-- model/
|-- resources/
|-- train.py
|-- requirements.txt
|-- .gitignore
|-- LICENSE
|-- README.md

Working Record

State	Network	Train Epochs / Time	Best Test Acc	Model Param
☑ 09 June, 2022	LeNet	30 / 18m-29s	0.9016	LeNet-best.pth.tar
☑ 11 June, 2022	AlexNet1	30 / 35m-39s	0.9219	AlexNet-best.pth.tar
☑ 13 June, 2022	VGGNet2	30 / 30m-35s	0.9135	VGGNet-best.pth.tar
☑ 22 June, 2022	InceptionNet	30 / 416m-54s	0.9274	InceptionNet-best.pth.tar
☑ 16 June, 2022	ResNet	30 / 82m-24s	0.9340	ResNet-best.pth.tar

Notes

• Device on GPU: NVIDIA GeForce GTX 1070, CPU: Intel i7-7700K, RAM: 32GB and Win10 System.
• Training loss curves can be seen in Loss Curves folder.
• ¹ Considering the image size of Fashion-MNIST, here in AlexNet has some tiny differences with the original AlexNet Framework, and we don't take the seperate group and LRN structure as well.
• ² Considering the image size of Fashion-MNIST, here in VGGNet16, we delete the last two block of convolutional layers, which are layer eight to thirteen.

References

1. LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
2. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25.
3. Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
4. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).
5. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).