Saliency Maps

Input-Gradient

Explainability method that uses the gradient of the output with respect to the input as a direct measure of relevance for each pixel or feature.

← Tillbaka