A Real-Time Face Detector Based on an End-to-End CNN
Chenghao Zheng, Menglong Yang, Chengpeng Wang · 2017
Convolutional neural networks with deeply trained make a significant performance improvement in face detection. However, the major shortcomings, i.e. need of high computational cost and slow calculation, make the existing CNN-based face detectors impractical in many applications. In this paper, a real-time approach for face detection was proposed by utilizing a single end-to-end deep neural network with multi-scale feature maps, multi-scale prior aspect ratios as well as confidence rectification. Multi-scale feature maps overcome the difficulties of detecting small face, and meanwhile, multiscale prior aspect ratios reduce the computing cost and the confidence rectification, which is in line with the biological intuition and can further improve the detection rate. Evaluated on the public benchmark, FDDB, the proposed algorithm, gained a performance as good as the state-of-the-art CNN-based methods, however, with much faster speed.