Scale-Aware Face Detection

TL;DR

The paper introduces SAFD, a scale-aware face detection method that explicitly handles face size variations using CNNs, reducing computational costs while maintaining high detection accuracy.

Contribution

SAFD explicitly models face scale distribution with CNNs and guides image zooming, improving efficiency and accuracy over traditional multi-scale methods.

Findings

Over 99% of faces in AFW detected with less than two zooms.

SAFD outperforms existing methods on FDDB, MALF, and AFW datasets.

Significant reduction in computation compared to multi-scale testing.

Abstract

Convolutional neural network (CNN) based face detectors are inefficient in handling faces of diverse scales. They rely on either fitting a large single model to faces across a large scale range or multi-scale testing. Both are computationally expensive. We propose Scale-aware Face Detector (SAFD) to handle scale explicitly using CNN, and achieve better performance with less computation cost. Prior to detection, an efficient CNN predicts the scale distribution histogram of the faces. Then the scale histogram guides the zoom-in and zoom-out of the image. Since the faces will be approximately in uniform scale after zoom, they can be detected accurately even with much smaller CNN. Actually, more than 99% of the faces in AFW can be covered with less than two zooms per image. Extensive experiments on FDDB, MALF and AFW show advantages of SAFD.