T-TAME: Trainable Attention Mechanism for Explaining Convolutional Networks and Vision Transformers

T-TAME is a general methodology for explaining deep neural networks used in image classification tasks, providing high-quality explanation maps in a single forward pass.