Conference: Accepted at WACV'21

Paper: Arxiv, Blog

Authors: Ragav Sachdeva, Filipe R. Cordeiro, Vasileios Belagiannis, Ian Reid, Gustavo Carneiro

python Train_cifar.py --r [total_noise] --on [proportion_of_openset_noise] --data_path [path_to_cifar10] --noisy_dataset [cifar100/imagenet32] --noise_data_dir [path_to_cifar100/imagenet32]

Note: r is the same as ρ in the paper and on is the same as (1-ω).

Thanks to Li et al. for publishing their code for DivideMix. Our implementation is heavily based on their work.