Gradcam_visualization

GradCAM Visualization of Mass Estimation From Images

Implementation

The code provided herein is implemented in TF1.12 and compatible with Eager mode.

To run the code

Simply run the visualize_cam.py with specifying the following arguments directly in terminal. If you would like to visualize a paritcular image, make sure to place the image in this directory /images/input/

Args

• '-i', '--input_image', default='sample.bmp', type=str, help= '(Full name of the input image -- default set to sample.bmp'
• '-o', '--output_image', default='sample.png', type=str, help='Full name of output image (should be .png) -- default set to input_image.png '

Simplist use

This runs the cam with sample.bmp image located in /images/input/

• python3 visualize_cam.py

Output

This code produces an CAM image for input image in this directory /images/ouput

Acknowledgements

Part of the code was adapted from: TF session implementation here

Author

Muhammad K.A. Hamdan