Neural network-based face detection

TLDR

The authors develop a neural network‑based upright frontal face detection system and a straightforward procedure for aligning positive face examples for training. The system uses a retinally connected neural network that scans small windows, arbitrates among multiple networks for improved performance, collects negative examples through a bootstrap algorithm that adds false detections, and applies heuristics such as non‑overlap of faces to enhance accuracy. The approach eliminates the need for manual negative example selection and achieves detection and false‑positive rates comparable to several state‑of‑the‑art face detection systems.

Abstract

We present a neural network-based upright frontal face detection system. A retinally connected neural network examines small windows of an image and decides whether each window contains a face. The system arbitrates between multiple networks to improve performance over a single network. We present a straightforward procedure for aligning positive face examples for training. To collect negative examples, we use a bootstrap algorithm, which adds false detections into the training set as training progresses. This eliminates the difficult task of manually selecting nonface training examples, which must be chosen to span the entire space of nonface images. Simple heuristics, such as using the fact that faces rarely overlap in images, can further improve the accuracy. Comparisons with several other state-of-the-art face detection systems are presented, showing that our system has comparable performance in terms of detection and false-positive rates.