This is a pytorch implemention of the following work Contrastive layerwise relevance propagation (CLRP):
Requirements: numpy==1.14.2; python==3.6.4; pytorch==1.1.0;
The code creates CLRP saliency maps to explain individual classification on VGG16 model.
python CLRP/run.py
The results are visualized as follows (LRP vs. CLRP):
If this repo is helpful for you, please cite our work.
@inproceedings{gu2018understanding,
title={Understanding individual decisions of cnns via contrastive backpropagation},
author={Gu, Jindong and Yang, Yinchong and Tresp, Volker},
booktitle={Asian Conference on Computer Vision},
pages={119--134},
year={2018},
organization={Springer}
}