/grad_cam

Primary LanguageJupyter Notebook

Grad-CAM and CAM implementation with Keras

This repo contains Keras implementation of CAM, Grad-CAM, Guided Grad-CAM on the 2017 Korean presidential election candidates' face image data.

CAM, Grad-CAM, Guided Grad-CAM

Original Paper links

CAM: https://arxiv.org/abs/1512.04150
Grad-CAM: https://arxiv.org/abs/1610.02391
Guided Backpropagation: https://arxiv.org/abs/1412.6806

Data Source

I crawled the original images from Google and cropped faces myself. The dataset is available for any research purposes. https://www.floydhub.com/junkwhinger/datasets/presidential_candidates_2017_v2

Related Blogposts (in Korean)

Transfer Learning: 대선주자 얼굴 분류기

http://jsideas.net/python/2017/05/07/transfer_learning.html

Transfer Learning wtih Keras on FloydHub

http://jsideas.net/python/2017/11/26/transfer_learning_with_keras_on_floydhub.html

CAM: 대선주자 얼굴 위치 추적기

http://jsideas.net/python/2018/01/04/class_activation_map.html

Grad-CAM: 대선주자 얼굴 위치 추적기

http://jsideas.net/python/2018/01/12/grad_cam.html

Acknowledgement

Keras implementation of class activation mapping

https://github.com/jacobgil/keras-cam

Grad-CAM implementation in Keras

https://github.com/jacobgil/keras-grad-cam#grad-cam-implementation-in-keras

Grad-CAM-tensorflow

https://github.com/insikk/Grad-CAM-tensorflow