Deep Learning CNN visualization in MATLAB using occlusion sensitivity and Grad-CAM

This video shows how to do visualization on a convolutional neural network (CNN) using the occlusionSensitivity command and the Grad-CAM technique (Gradient class activation mapping). These techniques help debugging CNN's.

References:
Grad-CAM Reveals the Why Behind Deep Learning Decisions

Understand Network Predictions Using Occlusion

Other visualization techniques: