Learning a discriminative feature for object detection based on feature fusing and context learning

Object detection is one of the most challenging tasks in the field of computer vision. It is widely used in traffic sign detection[1], pedestrian detection[2,3], person re-identification[4], object tracking[5,6,7] and so on[8,9]. Although convolutional neural network (CNN)-based algorithms have made great achievements in this field, object detection still suffers from illumination changes, occlusion, intraclass differences, etc.[10]. Candidate bounding box generation methods and feature extraction methods also influence the final detection result. In this paper, we propose a discriminative feature extraction method based on feature fusion and context learning.