An Efficient Method for Face Quality Assessment on the Edge

TL;DR

This paper introduces a low-cost face quality assessment method integrated into an edge device face recognition pipeline, improving detection prioritization with minimal additional computation.

Contribution

It presents a novel face quality scoring approach by adding a single layer to a landmark detection network, optimized for edge devices with surveillance-like data augmentation.

Findings

Outperforms state-of-the-art face quality regression models

Efficient implementation on edge GPUs

Effective in real-life surveillance scenarios

Abstract

Face recognition applications in practice are composed of two main steps: face detection and feature extraction. In a sole vision-based solution, the first step generates multiple detection for a single identity by ingesting a camera stream. A practical approach on edge devices should prioritize these detection of identities according to their conformity to recognition. In this perspective, we propose a face quality score regression by just appending a single layer to a face landmark detection network. With almost no additional cost, face quality scores are obtained by training this single layer to regress recognition scores with surveillance like augmentations. We implemented the proposed approach on edge GPUs with all face detection pipeline steps, including detection, tracking, and alignment. Comprehensive experiments show the proposed approach's efficiency through comparison with SOTA face quality regression models on different data sets and real-life scenarios.