Grad-CAM
An explainability technique that uses gradients flowing into the final convolutional layer to produce a heat map highlighting important regions of an image.
Think of It Like This
Like eye-tracking software that records exactly where a person is looking on a screen to understand why they clicked on a specific advertisement.
Gradient-weighted Class Activation Mapping (Grad-CAM) helps visualize the decision-making process of Convolutional Neural Networks. By projecting the class-specific gradients back onto the input image, it shows which pixels most heavily influenced the final prediction. This is widely used in medical imaging to verify that models are looking at actual pathologies.