GradCAM

Explore the GradCAM technique to visualize important regions within images influencing deep learning model predictions. Understand how gradients and feature maps interact to generate class activation maps, helping interpret model decisions. This lesson covers practical implementation steps using MobileNet-V2 and ImageNet-1K to produce heat maps that highlight key features driving classification outcomes.

We'll cover the following...

Gradient-weighted CAM
Implementation

Gradient-weighted CAM

A gradient-weighted CAM or Grad-CAM is another class activation map algorithm that uses the gradients of the predicted class propagating into the final convolutional layer to weigh the feature maps used for creating maps.

Let's assume that $\mathcal{F}_l \in \R^{h \times w}$ is the $l^{th}$ feature map obtained from the penultimate convolutional layer ( $h$ and $w$ are the height and width of the feature map) and $\mathcal{F}_l(i,j)$ represents the activation of the $(i, j)^{t h}$ ...

1.Introduction to Explainable AI

Project

2.Saliency Maps

3.Class Activation Maps

4.Miscellaneous Methods

5.Metrics of Interpretability

Assessment

Mini Project

GradCAM

Gradient-weighted CAM