An Explainable Model-Agnostic Algorithm for CNN-based Biometrics Verification

TL;DR

This paper adapts the LIME explanation method for CNN-based biometric verification, focusing on face recognition, by using cosine similarity of feature vectors instead of softmax probabilities, enabling interpretability in unseen classes.

Contribution

It introduces a novel adaptation of LIME for biometric verification that operates with feature vectors and cosine similarity, addressing the challenge of unseen classes during explanation.

Findings

Effective explanations for face verification models

Applicable to MobileNetv2 and ResNet50 architectures

Enhances interpretability in biometric verification

Abstract

This paper describes an adaptation of the Local Interpretable Model-Agnostic Explanations (LIME) AI method to operate under a biometric verification setting. LIME was initially proposed for networks with the same output classes used for training, and it employs the softmax probability to determine which regions of the image contribute the most to classification. However, in a verification setting, the classes to be recognized have not been seen during training. In addition, instead of using the softmax output, face descriptors are usually obtained from a layer before the classification layer. The model is adapted to achieve explainability via cosine similarity between feature vectors of perturbated versions of the input image. The method is showcased for face biometrics with two CNN models based on MobileNetv2 and ResNet50.