Scale-Aware Face Detection

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.

Accepted to CVPR 2017

References in corpus (5)

• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• DenseBox: Unifying Landmark Localization with End to End Object Detection
• From Facial Parts Responses to Face Detection: A Deep Learning Approach
• Occlusion Coherence: Detecting and Localizing Occluded Faces
• Convolutional Channel Features

Cited by in corpus (18)

• Face Attention Network: An Effective Face Detector for the Occluded Faces
• Face Detection Using Improved Faster RCNN
• Registration-free Face-SSD: Single shot analysis of smiles, facial attributes, and affect in the wild
• Dynamic Neural Networks: A Survey
• Anchor Cascade for Efficient Face Detection
• FSRNet: End-to-End Learning Face Super-Resolution with Facial Priors
• iQIYI-VID: A Large Dataset for Multi-modal Person Identification
• Accurate Face Detection for High Performance
• Robust and High Performance Face Detector
• Selective Refinement Network for High Performance Face Detection
• SFace: An Efficient Network for Face Detection in Large Scale Variations
• MSR: Multi-Scale Shape Regression for Scene Text Detection
• Recurrent Scale Approximation for Object Detection in CNN
• Exploring Multi-Branch and High-Level Semantic Networks for Improving Pedestrian Detection
• Beyond Trade-off: Accelerate FCN-based Face Detector with Higher Accuracy
• SurfConv: Bridging 3D and 2D Convolution for RGBD Images
• Instance Scale Normalization for image understanding
• KPNet: Towards Minimal Face Detector