Grad-CAM: Why did you say that?
Abstract
We propose a technique for making Convolutional Neural Network (CNN)-based models more transparent by visualizing input regions that are 'important' for predictions -- or visual explanations. Our approach, called Gradient-weighted Class Activation Mapping (Grad-CAM), uses class-specific gradient information to localize important regions. These localizations are combined with existing pixel-space visualizations to create a novel high-resolution and class-discriminative visualization called Guided Grad-CAM. These methods help better understand CNN-based models, including image captioning and visual question answering (VQA) models. We evaluate our visual explanations by measuring their ability to discriminate between classes, to inspire trust in humans, and their correlation with occlusion maps. Grad-CAM provides a new way to understand CNN-based models. We have released code, an online demo hosted on CloudCV, and a full version of this extended abstract.
- Publication:
-
arXiv e-prints
- Pub Date:
- November 2016
- DOI:
- 10.48550/arXiv.1611.07450
- arXiv:
- arXiv:1611.07450
- Bibcode:
- 2016arXiv161107450S
- Keywords:
-
- Statistics - Machine Learning;
- Computer Science - Computer Vision and Pattern Recognition;
- Computer Science - Machine Learning
- E-Print:
- Presented at NIPS 2016 Workshop on Interpretable Machine Learning in Complex Systems. This is an extended abstract version of arXiv:1610.02391 (CVPR format)