The unofficial reimplementation of fixmatch with RandomAugment.

2020-03-30_18:07:08.log : annotation decay and add classifier.bias

2020-03-31_09:51:38.log : add interleave and run model once

• python 3.6
• pytorch 1.3.1
• torchvision 0.2.1

The other packages and versions are listed in requirements.txt. You can install them by pip install -r requirements.txt.

download cifar-10 dataset:

    $ mkdir -p dataset && cd data
    $ wget -c http://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
    $ tar -xzvf cifar-10-python.tar.gz

download cifar-100 dataset:

    $ mkdir -p dataset && cd data
    $ wget -c http://www.cs.toronto.edu/~kriz/cifar-100-python.tar.gz
    $ tar -xzvf cifar-100-python.tar.gz

To train the model on CIFAR10 with 40 labeled samples, you can run the script:

    $ CUDA_VISIBLE_DEVICES='0' python train.py --dataset CIFAR10 --n-labeled 40 

To train the model on CIFAR100 with 400 labeled samples, you can run the script:

    $ CUDA_VISIBLE_DEVICES='0' python train.py --dataset CIFAR100 --n-labeled 400 

• https://github.com/CoinCheung/fixmatch
• https://github.com/kekmodel/FixMatch-pytorch
• official implement https://github.com/google-research/fixmatch