Regional attention reinforcement learning for rapid object detection. (March 2022)

• Record Type: Journal Article
• Authors: Yao, Hongge; Dong, Peng; Cheng, Siyi; Yu, Jun
• Is Part Of: Computers & electrical engineering. Volume 98(2022)
• Publication Date: 2022-03
• Subjects: Regional attention -- Reinforcement learning -- Object detection -- Information fusion -- Location and recognition; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.107747
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: When people observe a picture, they first pay attention to local areas of the picture, rather than the whole areas, then combine them with previous experience in the brain, and finally make judgments through thinking. This is human visual logic. In this paper, we propose a regional attention reinforcement learning model for object detection. The proposed model uses human visual logical to solve the detection problem of small and complex targets in the picture. The model uses a recurrent network structure as the main framework to extract historical information, and fuse the historical information with the current concerned information. At each recurrent time step, it can pay attention to the fused information, especially pay more attention to the information that may have objects. Experimental results show that the proposed method has more than 5% improved in recognition accuracy to the conventional methods. In terms of FLOPs, the conventional methods normally require 170 M, while the proposed method only needs 25.4M This means that the proposed method has higher detection efficiency.