IDAA

Official implementation:

• Identity-Disentangled Adversarial Augmentation for Self-Supervised Learning, ICML 2022. (Paper)

Architecture and pipeline of Identity-Disentangled Adversarial Augmentation (IDAA)

For questions, you can contact (kwyang@mail.ustc.edu.cn).

Requirements

1. Python
2. Pytorch
3. Wandb
4. Torchvision
5. Apex(optional)

Pretrain a VAE

python train_vae.py --dim 512 --kl 0.1 --save_dir ./results/vae_cifar10_dim512_kl0.1_simclr --mode simclr --dataset cifar10

Apply IDAA to SimCLR

cd SimCLR

SimCLR training and evaluation:

python main.py --seed 1 --gpu 0  --dataset cifar10 --resnet resnet18;
python eval_lr.py --seed 1 --gpu 0 --dataset cifar10 --resnet resnet18

SimCLR+IDAA training and evaluation:

python main.py --adv --eps 0.1 --seed 1 --gpu 0 --dataset cifar10 --dim 512 --vae_path ../results/vae_cifar10_dim512_kl0.1_simclr/model_epoch292.pth --resnet resnet18;
python eval_lr.py --adv --eps 0.1 --seed 1 --gpu 0 --dataset cifar10 --dim 512 --resnet resnet18

References

We borrow some code from https://github.com/chihhuiho/CLAE.

Citation

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

@inproceedings{yang2022identity,
  title={Identity-Disentangled Adversarial Augmentation for Self-supervised Learning},
  author={Yang, Kaiwen and Zhou, Tianyi and Tian, Xinmei and Tao, Dacheng},
  booktitle={International Conference on Machine Learning},
  pages={25364--25381},
  year={2022},
  organization={PMLR}
}