Gradient class activation maps are a visualization technique for deep learning networks. This is my implementation of Grad-CAM in tensorflow. I rewrite Grad-CAM funcion through it is a part of VGG graph. CAM may be obtained directly by launching graph.
The original paper: https://arxiv.org/pdf/1610.02391v1.pdf
Download the VGG16 weights from https://www.cs.toronto.edu/~frossard/vgg16/vgg16_weights.npz and put it to project directory.
python3 main.py --input laska.png --output laska_save.png| Input | Output |
|---|---|
 |
 |
All code was taken from https://github.com/Ankush96/grad-cam.tensorflow.git and modified a little.