Pedestrian detection based on YOLOv2 with skip structure in underground coal mine

Abstract

Pedestrian detection is an important topic in object detection. Compared with other object detectors, YOLOv2 achieves high accuracy and fast speed for general object detection, however it degrades accuracy when detecting crowed pedestrians. In this paper, combining with the skip structure of FCN, we tailor the YOLOv2 network to improve the accuracy in detecting small pedestrians which appear in groups in underground coal mine. In this way, we propose two modified versions of YOLOv2 which are YWSSv1 and YWSSv2. Compared with YOLOv2, YWSSv1 slightly improves 0.1 mAP but keeps the same fast speed. YWSSv2 significantly gains 12 mAP higher than YOLOv2 but sacrifices its speed at just 5 FPS.