Wide aspect ratio matching for robust face detection

Abstract

Recently, anchor-based methods have achieved great progress in face detection. They adopt standard anchor matching strategy to sample positive anchors according to predefined IoU threshold. However, the max IoUs of extreme aspect ratio faces are still lower than fixed positive threshold, leading to the sampling failure from these faces. To construct a more robust detection model, more positive anchors from extreme aspect ratio faces need to be sampled and participate in the training phase. The goal of the present research is to improve the detection performance by reasonably extending sampling range of face aspect ratio. In this paper, we firstly explore the factors that affect the max IoU of each face in theory. Then, anchor matching simulation is performed to evaluate the sampling range of face aspect ratio. Finally, we propose a Wide Aspect Ratio Matching (WARM) strategy to collect more representative positive anchors from ground-truth faces across a wide range of aspect ratios. Besides, we present a novel feature enhancement module, named Receptive Field Diversity (RFD) module, to provide diverse receptive field corresponding to different aspect ratios. Extensive experiments have been conducted on popular benchmarks to show the effectiveness of our method, which can help detectors better capture extreme aspect ratio faces. Our method achieves promising APs on WIDER FACE validation dataset (easy: 0.965, medium: 0.955, hard: 0.904) and impressive generalization capability on FDDB dataset.

An open dataset for intelligent recognition and classification of abnormal condition in longwall mining

“…Currently, datasets are widely used in automatic driving, object detection, face recognition, natural language processing, text detection, medical and other fields [7][8][9][10] . Some widely used object detection datasets are as follows: (1) COCO datasets with large-scale commonly used items as target detection objects [11][12][13] ; (2) VOC datasets with people, common animals, traffic vehicles, indoor furniture objects as target detection objects [14][15][16] ; (3) DOTA dataset with airplanes, ships, storage tanks, baseball stadiums, tennis courts, basketball courts, ground runways, ports, bridge as target detection objects [17][18][19] ; (4) TT100K dataset with common vehicles as the target detection object [20][21][22] ; (5) WIDER FACE dataset with facial expression, illumination and posture as target detection objects [23][24][25] ; (6) YOLO format dataset that dedicated to the target detection [26][27][28] , etc. In addition to these common datasets, we can also customize the dataset through pytorch framework, but the custom dataset format is complex, diversified and poor sharing 29 .…”

A convolutional neural network for face mask detection in IoT-based smart healthcare systems

“…There is insufficient cultural opinion on whether to force people to wear face masks against transmission of corona virus. The researchers (Luo et al, 2022) demonstrates the causes of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and analyses the importance of wearing the mask. During pandemic situations, people are advised to use mask and maintain social distancing (Luo et al, 2022).…”

“…The researchers (Luo et al, 2022) demonstrates the causes of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and analyses the importance of wearing the mask. During pandemic situations, people are advised to use mask and maintain social distancing (Luo et al, 2022). Because of the scarcity of the training samples, most of the above models suffered overfitting problem.…”