Human interpretability of AI-mediated comparisons of sparse natural person photos

While AI can be of great use in situations where many images of an individual exist, it is of limited assistance in determining the identity of individuals in the case of sparse photographs (2-3) of an individual. Furthermore, it is difficult for humans to understand how a decision of matching or not-matching is reached by an AI agent. This problem is recognized by the ONI/DST Artificial Intelligence for Decision Making Initiative Task 2 - Facial Verification Explainability Measures, which this proposal addresses. I will use interpretable facial feature ratio comparison as a human-friendly way of explaining photo verification of natural persons in photographs.