PEAL: PyTorch Explain and Adapt Library
PEAL is a PyTorch library for generating visual counterfactual explanations and for correcting classifiers and foundation-model probes with them. It contains the official implementations of Smoothed Counterfactual Explorer (SCE), Counterfactual Knowledge Distillation (CFKD) and Disentangled Diffusion Autoencoders (DiDA...