QPDet: A Queuing People Detector for Aerial Images.

Object detection in aerial images is a fundamental and active research topic with astonishing achievements in recent years. However, tiny object detection in aerial images remains a challenging problem due to the small sizes and limited features of the objects. In this letter, we propose QPDet, a queuing people (QP) detector for aerial images. First, we create a large-scale dataset of QP using unmanned aerial vehicle (UAV). Second, a new label assigner is proposed based on the normally distributed receptive field. Third, a random subimage segmentation and augmentation (RISA) strategy is developed to facilitate the training of different queuing patterns. Our method is simple, yet effective. Extensive experiments have been conducted to verify the effectiveness of our model, where we outperform most of the state-of-the-art methods in accuracy.