Hourglass Face Detector for Hard Face

Zijun Yu,Jian Yin,Qiang Zhang,Wenming Yang,Jing-Hao Xue,Qingmin Liao

DOI: https://doi.org/10.1109/ijcnn52387.2021.9533674

2021-01-01

Abstract:Face detection is an upstream task of facial image analysis. In many real-world scenarios, we need to detect small, occluded or dense faces that are hard to detect, but hard face detection is a challenging task in particular considering the balance between accuracy and inference speed for real-world applications. This paper proposes an Hourglass Face Detector (HFD) for hard face by developing a deep one-stage fully-convolutional hourglass network, which achieves an excellent balance between accuracy and inference speed. To this end, the HFD firstly shrinks a feature map by a series of stridden convolutional layers rather than pooling layers, so that useful subtle information is preserved better. Secondly, it exploits context information by merging fine-grained shallow feature maps with deep ones full of semantic information, making a better fusion of detailed information and semantic information to achieve a better detection of small faces. Moreover, the HFD exploits prior and multiscale information from the training data to enhance its scale-invariance and adaptability of anchor scales. Compared with the SSH and S3FD methods, the HFD can achieve a better performance in average precision on detecting hard faces as well as a quicker inference. Experiments on the WIDER FACE and FDDB datasets demonstrate the superior performance of our proposed method.