ExplainLFS: Explaining neural architectures for similarity learning from local perturbations in the latent feature space

Marilyn Bello, Pablo Costa, Gonzalo Nápoles, Pablo Mesejo, Oscar Cordón

Abstract: Highlights•An explanation technique for similarity learning neural architectures is presented.•Our approach establishes a visual connection between the embedding and the image pixels.•It aims to identify relevant features in similarity learning.•The method’s effectiveness is demonstrated in face recognition and image retrieval.