Grad-CAM

A widely used explainable AI technique, introduced by Selvaraju et al. in 2017, that produces a class-discriminative heat map over an input image by weighting convolutional feature maps by the gradient of the target class score. Grad-CAM and its variants (SmoothGrad-CAM, Score-CAM, Layer-CAM) are the default visual explanation tool for image-recognition models and are often deployed alongside assistive apps for blind and low-vision users. Ironically, because Grad-CAM's output is itself a visual heat map, it exemplifies the perceptual gap: the explanation is inaccessible to the very users who cannot see the original image.