Official Pytorch implementation for Addressing out-of-distribution label noise in webly-labelled data.

Paper link: thecvf

Set dataset location

The path to the datasets should be set in the mypath.py file

Experiments

bash train.sh

You can resume a checkpoint using

--resume path/to/checkpoint.pth.tar

Trained model

The models trained on Webvision can be downloaded here (--seeds 1, 2 ,3): gdrive

Testing on imagenet

The model trained on Webvision can be tested on the Imagenet validation set using

python eval_imagenet.py model1_state_dict.pth.tar

and optionally for an ensemble

python eval_imagenet.py model1_state_dict.pth.tar model2_state_dict.pth.tar

Cite the paper

@inproceedings{2022_WACV_DSOS,
  title="{Addressing out-of-distribution label noise in webly-labelled data}",
  author="Albert, Paul and Ortego, Diego and Arazo, Eric and O{'}Connor, Noel and McGuinness, Kevin",
  booktitle="{Winter Conference on Applications of Computer Vision (WACV)}",
  year="2022"}