liz6688 / DiLo

Code and pertained models for the paper "Distilling Localization for Self-Supervised Representation Learning"

DiLo Code and Pretrained Models

Distilling Localization for Self-Supervised Representation Learning
Nanxuan Zhao* Zhirong Wu* Rynson W.H. Lau Stephen Lin

Dataloader

saliency_dataset.py is an example dataloader for adding the DiLo copy-and-paste augmentation.

The saliency estimation models can be found:

• GS: paper , code
• MC: paper , code
• RBD: paper , code
• BASNet: code , model pretrained from scratch trained on DUTS dataset for 150 epochs.

Pretrained Models

Base Model	Saliency Estimation Model	download
MoCo	RBD	model
MoCo	BASNet	model
MoCo v2	RBD	model
MoCo v2	BASNet	model

Citation

Please cite our paper if you use DiLo in your research or wish to refer to the results published in the paper.

@inproceedings{ZhaoAAAI2021, 
    author = {Nanxuan Zhao and Zhirong Wu and Rynson W.H. Lau and Stephen Lin}, 
    title = {Distilling Localization for Self-Supervised Representation Learning}, 
    booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence}, 
    year = {2021} 
}