Home

Awesome

Unsupervised Feature Learning via Non-parameteric Instance Discrimination

This repo constains the pytorch implementation for the CVPR2018 unsupervised learning paper (arxiv).

Updated Pretrained Model

An updated instance discrimination model with memory bank implementation and with nce-k=65536 negatives is provided. The updated model is trained with Softmax-CE loss as in CPC/MoCo instead of the original NCE loss.

Oldies: original releases of ResNet18 and ResNet50 trained with 4096 negatives and the NCE loss. Each tar ball contains the feature representation of all ImageNet training images (600 mb) and model weights (100-200mb). You can also get these representations by forwarding the network for the entire ImageNet images.

Highlight

Nearest Neighbor

Please follow this link for a list of nearest neighbors on ImageNet. Results are visualized from our ResNet50 model, compared with raw image features and supervised features. First column is the query image, followed by 20 retrievals ranked by the similarity.

Usage

Our code extends the pytorch implementation of imagenet classification in official pytorch release. Please refer to the official repo for details of data preparation and hardware configurations.

Citation

@inproceedings{wu2018unsupervised,
  title={Unsupervised Feature Learning via Non-Parametric Instance Discrimination},
  author={Wu, Zhirong and Xiong, Yuanjun and Stella, X Yu and Lin, Dahua},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2018}
}

Contact

For any questions, please feel free to reach

Zhirong Wu: xavibrowu@gmail.com