Facial Image Feature Analysis and its Specialization for Fr\'echet Distance and Neighborhoods

TL;DR

This paper investigates how domain-specific feature training affects the Fréchet distance and neighborhood analysis in facial images, highlighting the importance of training data domain in image distance metrics.

Contribution

It provides the first analysis of domain-specific feature training effects on Fréchet distance and neighborhoods in facial images, supported by extensive experiments and user studies.

Findings

Domain-specific training improves distance measurement accuracy.

Self-supervised learning enhances feature relevance for facial images.

User studies confirm the practical impact of domain adaptation.

Abstract

Assessing distances between images and image datasets is a fundamental task in vision-based research. It is a challenging open problem in the literature and despite the criticism it receives, the most ubiquitous method remains the Fr\'echet Inception Distance. The Inception network is trained on a specific labeled dataset, ImageNet, which has caused the core of its criticism in the most recent research. Improvements were shown by moving to self-supervision learning over ImageNet, leaving the training data domain as an open question. We make that last leap and provide the first analysis on domain-specific feature training and its effects on feature distance, on the widely-researched facial image domain. We provide our findings and insights on this domain specialization for Fr\'echet distance and image neighborhoods, supported by extensive experiments and in-depth user studies.