Part-based Face Recognition with Vision Transformers

TL;DR

This paper introduces a novel part-based face recognition pipeline using Vision Transformers, achieving state-of-the-art accuracy by combining landmark prediction and patch-based processing without landmark supervision.

Contribution

It presents a new part-based face recognition method with Vision Transformers that surpasses existing methods and operates end-to-end without landmark annotations.

Findings

fViT surpasses most state-of-the-art face recognition methods.

Part fViT further improves accuracy by learning discriminative patches.

The approach achieves state-of-the-art results on multiple benchmarks.

Abstract

Holistic methods using CNNs and margin-based losses have dominated research on face recognition. In this work, we depart from this setting in two ways: (a) we employ the Vision Transformer as an architecture for training a very strong baseline for face recognition, simply called fViT, which already surpasses most state-of-the-art face recognition methods. (b) Secondly, we capitalize on the Transformer's inherent property to process information (visual tokens) extracted from irregular grids to devise a pipeline for face recognition which is reminiscent of part-based face recognition methods. Our pipeline, called part fViT, simply comprises a lightweight network to predict the coordinates of facial landmarks followed by the Vision Transformer operating on patches extracted from the predicted landmarks, and it is trained end-to-end with no landmark supervision. By learning to extract discriminative patches, our part-based Transformer further boosts the accuracy of our Vision Transformer baseline achieving state-of-the-art accuracy on several face recognition benchmarks.