Multi‐scale pedestrian detection with global–local attention and multi‐scale receptive field context

As a basic component in the field of computer vision, pedestrian detection plays an essential role several real-world applications such as video surveillance. The promising performance has been achieved relying on deep learning, but large-scale variance and small-scale remain inherently hard before. In order to deal with aforementioned problems, this paper proposes multi-scale method global–local attention receptive context (MRFC). To make network focus pedestrians, we add high-resolution branch original detector. better integrate incongruous semantic feature, module is embedded highlight feature representation pedestrians so implement fusion effectively. adapt achieve scale-variance detection, MRFC applied. Based integrating above structures, proposed achieves competitive results Caltech CityPersons datasets. source code released https://github.com/xiaopan999/yolov5-pedestrian_detection.