Hourglass Face Detector for Hard Face

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 fullyconvolutional 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 S 3 FD 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.

An Evaluation of Deep Learning-Based Object Identification

“…An SSH was created by agghey and colleagues [94] in order to enable multi-scale face recognition by performing detection on feature maps of varying sizes. Anchor approaches such as FaceBoxes [95], S3FD [96], and ScaleFace [97] are also able to handle the detection of tiny objects and multi-scale faces. Salient object detection: Bringing attention to the principal object areas in an image, which are often referred to as the salient regions, is the objective of the technique of salient object detection.…”