EigenCAM
A white-box explainability method that generates saliency maps by applying principal component analysis to convolutional neural network feature maps. It requires no input perturbation, is computationally efficient, and is applicable to model interpretation for object detection and image classification tasks.
Timeline (last 90 days)
EigenCAM is a white-box gradient/activation map method that generates saliency maps using the principal components of convolutional layer feature maps without perturbing the input, making it a non-perturbation-based method.
EigenCAM's overall faithfulness is lower than D-RISE, but it similarly exhibits score coupling.
All Facts (2)
EigenCAM is a white-box gradient/activation map method that generates saliency maps using the principal components of convolutional layer feature maps without perturbing the input, making it a non-perturbation-based method.
50%UnverifiedEigenCAM's overall faithfulness is lower than D-RISE, but it similarly exhibits score coupling.
50%