Explain output of image classifier

VGG-16 [1] trained on ILSVRC-2014 data [2]. The algorithm used for identifying the sub-images that gets the highest probability for the top class is inspired by [3].

Weights for VGG16 can be downloaded here: vgg16_weights.h

Usage

python explain.py -i dog.jpg -m vgg16_weights.h5 -o out.jpg

Input image:

Output image:

[1] Very Deep Convolutional Networks for Large-Scale Image Recognition K. Simonyan, A. Zisserman arXiv:1409.1556

[2] http://image-net.org/challenges/LSVRC/2014/

[3] "Why Should I Trust You?": Explaining the Predictions of Any Classifier Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin arXiv:1602.04938