Open-sampling

This repository is the official implementation of Open-sampling (ICML 2022: Open-Sampling: Exploring Out-of-Distribution Data for Re-balancing Long-tailed Datasets).

Requirements

To install requirements:

pip install -r requirements.txt

Training

To train the model(s) in the paper, run this command:

python main.py --dataset cifar10 --gpu 0 --imb_type exp --imb_factor 0.01 --alg open -p 100 --lambda_o 1 -ab 512

Evaluation

We also provide a pre-trained ResNet-32 model in ./checkpoint/cifar10_open_CE_exp_0.01_1_1/ckpt.best.pth.tar, and the training log is in ./log/cifar10_open_CE_exp_0.01_1_1/

To evaluate the pre-trained model on CIFAR-10, run:

python test.py --dataset cifar10 --gpu 0 --resume ./checkpoint/cifar10_open_CE_exp_0.01_1_1/ckpt.best.pth.tar

What's More?

Below are my other research works related to this topic:

1. Using OOD examples to improve robustness against inherent noisy labels: NeurIPS 2021 | Code
2. Improving OOD detection by logit normalization: ICML 22 | Code

Citation

If you find this useful in your research, please consider citing:

@inproceedings{wei@open,
  title={Open-Sampling: Exploring out-of-distribution data for re-balancing long-tailed datasets}, 
  author={Wei, Hongxin and Tao, Lue and Xie, Renchunzi and Feng, Lei and An, Bo},
  booktitle = {International Conference on Machine Learning (ICML)},
 	year = {2022},
	organization={PMLR}
}