GradCAM

A PyTorch implementation of Grad-CAM based on ICCV 2017 paper Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization.

Requirements

• Anaconda
• PyTorch

  conda install pytorch torchvision -c pytorch

• opencv

  conda install opencv

Usage

python vis_grad_cam.py --image_name dog_and_cat.png --save_name result.png
optional arguments:
--image_name              the tested image name [default value is both.png]
--save_name               saved image name [default value is grad_cam.png]

Result