Beyond Trade-off: Accelerate FCN-based Face Detector with Higher Accuracy

Abstract

Fully convolutional neural network (FCN) has been dominating the game of face detection task for a few years with its congenital capability of sliding-window-searching with shared kernels, which boiled down all the redundant calculation, and most recent state-of-the-art methods such as Faster-RCNN, SSD, YOLO and FPN use FCN as their backbone. So here comes one question: Can we find a universal strategy to further accelerate FCN with higher accuracy, so could accelerate all the recent FCN-based methods? To analyze this, we decompose the face searching space into two orthogonal directions, `scale' and `spatial'. Only a few coordinates in the space expanded by the two base vectors indicate foreground. So if FCN could ignore most of the other points, the searching space and false alarm should be significantly boiled down. Based on this philosophy, a novel method named scale estimation and spatial attention proposal ($S^{2}AP$) is proposed to pay attention to some specific scales and valid locations in the image pyramid. Furthermore, we adopt a masked-convolution operation based on the attention result to accelerate FCN calculation. Experiments show that FCN-based method RPN can be accelerated by about $4\times$ with the help of $S^{2}AP$ and masked-FCN and at the same time it can also achieve the state-of-the-art on FDDB, AFW and MALF face detection benchmarks as well.

arXiv commentAccepted by CVPR2018

References in corpus (15)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Deep Residual Learning for Image Recognition
• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• Going Deeper with Convolutions
• Feature Pyramid Networks for Object Detection
• DenseBox: Unifying Landmark Localization with End to End Object Detection
• Rethinking Feature Discrimination and Polymerization for Large-scale Recognition
• From Facial Parts Responses to Face Detection: A Deep Learning Approach
• Quality Aware Network for Set to Set Recognition
• Style Aggregated Network for Facial Landmark Detection
• Face Detection with the Faster R-CNN
• Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
• Scale-Aware Face Detection
• Face Detection using Deep Learning: An Improved Faster RCNN Approach
• Exploring Disentangled Feature Representation Beyond Face Identification