Fashion-MNIST is a dataset of Zalando's article images—consisting of a training set of 60,000 examples and a test set of 10,000 examples. Each example is a 28x28 grayscale image, associated with a label from 10 classes. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms. It shares the same image size and structure of training and testing splits.
In this repo, I try to do some experiments with Fashion-MNIST dataset. (I do some experiments with Cifar-10 dataset either.)
- Standard preprocessing (mean/std subtraction/division) and data augment(Rand augment, random crops,horizontal flips, random erasing);
- Backbone: wide-resnet 40-4;
- Learning rate schedule:
CosineAnnealingLR; - Other tricks: Label smoothing, Exponential Moving Average, and so on.
1.run pip install -r requirements.txt;
2.choose a python script to run.
| script name | dataset | best test accuracy | weight files |
|---|---|---|---|
| fn_fmix_40_4_gn_ws_learning.py | Fashion-MNIST | 96.44%(epoch 562) | Google Cloud |
| fn_fmix_40_4_bn_mish_ws_gem.py | Fashion-MNIST | 96.69%(epoch 881) | Google Cloud |
| kaggle_cifar10_fmix_40_4_bn_mish_ws_gem.py | Cifar-10 | 98.03%(epoch 864) | Google Cloud |
| fn_96_74.py | Fashion-MNIST | 96.74%(epoch 785) | Google Cloud |