Home

Awesome

ACPL

This repository is the official implementation of ACPL: Anti-curriculum Pseudo-labelling for Semi-supervised Medical Image Classification . Structure

<!-- ![Framework](figs/framework.png) -->

Requirements

Datasets Preparation

Training

  1. 2% labelled set semi-supervised training:

    CUDA_VISIBLE_DEVICES=0,1 python main.py --data <data_dir> --task cx14 --resize 512 --batch-size 8 --epochs 20  --dist-url 'tcp://localhost:10001' --multiprocessing-distributed --world-size 1 --rank 0  --desc ACPL_2_1_k200 --num-workers 4 --eval-interval 100 --reinit --label_ratio 2 --runtime 1 --topk 200 --pl-epochs 10  --ds-mixup --sel 2 --num-gmm-sets 3 --lr 0.03  --resume <resume_from_ck>

    where the arguments represent:

    • data_dir - Chest X-ray14 root dir
    • desc - folder name for experiment description
    • label_ratio - labelled set size
    • runtime - multiple run
    • topk - KNN K
    • pl-epochs - train epochs for after assign pseudo labels
    • ds-mixup - use density mixup
    • sel - select high informative subset
    • num_gmm-sets - {low, medium, high} number of GMM sets
    • resume - resume from supervised training checkpoint, remove if training from scratch

  2. In origin paper we use DenseNet-121 as backbone and compute graph on GPU with Faiss. For larger backbone, You can switch to CPU in case GPU memory runout.

Citation

If you find this repo useful for your research, please consider citing our paper:

@article{@article{liu2021acpl,
  title={ACPL: Anti-curriculum Pseudo-labelling forSemi-supervised Medical Image Classification},
  author={Liu, Fengbei and Tian, Yu and Chen, Yuanhong and Liu, Yuyuan and Belagiannis, Vasileios and Carneiro, Gustavo},
  journal={arXiv preprint arXiv:2111.12918},
  year={2021}
}
<!-- # ACPL: Anti-curriculum Pseudo-labelling for Semi-supervised Medical Image Classification ## Introduction Pytorch code for *ACPL* paper ## Prerequisites Required packages to run: * Python 3.8 * Pytorch 1.10.0 * torchvision 0.11.1 * loguru 0.5.3 * faiss 1.7.1 The code is tested in two NVIDIA GTX 2080ti. Single card is supported but performance cannot guarantee. ## Quick run of the code 1. First, please download Chest X-ray14 from https://nihcc.app.box.com/v/ChestXray-NIHCC 2. Move train_val_list_2_1.txt to data directory 3. 2% labelled set semi-supervised training: ``` CUDA_VISIBLE_DEVICES=0,1 python main.py --data <data_dir> --task cx14 --resize 512 --batch-size 8 --epochs 20 --dist-url 'tcp://localhost:10001' --multiprocessing-distributed --world-size 1 --rank 0 --desc ACPL_2_1_k200 --num-workers 4 --eval-interval 100 --reinit --label_ratio 2 --runtime 1 --topk 200 --pl-epochs 10 --ds-mixup --sel 2 --num-gmm-sets 3 --lr 0.03 --resume <resume_from_ck> ``` where the arguments represent: * `data_dir` - Chest X-ray14 root dir * `desc` - folder name for experiment description * `label_ratio` - labelled set size * `runtime` - multiple run * `topk` - KNN K * `pl-epochs` - train epochs for after assign pseudo labels * `ds-mixup` - use density mixup * `sel` - select high informative subset * `num_gmm-sets` - {low, medium, high} number of GMM sets * `resume` - resume from supervised training checkpoint, remove if training from scratch -->