Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-Free Localization
Abstract
In response to recent criticism of gradient-based visualization techniques, we propose a new methodology to generate visual explanations for deep Convolutional Neural Networks (CNN) - based models. Our approach - Ablation-based Class Activation Mapping (Ablation CAM) uses ablation analysis to determine the importance (weights) of individual feature map units w.r.t. class. Further, this is used to produce a coarse localization map highlighting the important regions in the image for predicting the concept. Our objective and subjective evaluations show that this gradient-free approach works better than state-of-the-art Grad-CAM technique. Moreover, further experiments are carried out to show that Ablation-CAM is class discriminative as well as can be used to evaluate trust in a model.
Cite
Text
Desai and Ramaswamy. "Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-Free Localization." Winter Conference on Applications of Computer Vision, 2020.Markdown
[Desai and Ramaswamy. "Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-Free Localization." Winter Conference on Applications of Computer Vision, 2020.](https://mlanthology.org/wacv/2020/desai2020wacv-ablationcam/)BibTeX
@inproceedings{desai2020wacv-ablationcam,
title = {{Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-Free Localization}},
author = {Desai, Saurabh and Ramaswamy, Harish Guruprasad},
booktitle = {Winter Conference on Applications of Computer Vision},
year = {2020},
url = {https://mlanthology.org/wacv/2020/desai2020wacv-ablationcam/}
}