Candidate box fusion based approach to adjust position of the candidate box for object detection

• DOI: 10.1049/ipr2.12264
• Publication Date: 2021-06-24T10:47:01Z
• Authors: Jie Cao; Wei Ren; Hong Zhang; Zuohan Chen
• Keywords: Position (finance)

Abstract The method of object detection has been applied to all aspects in our lives. Although methods based on deep learning have widely used various fields, there are still some overlooked problems the candidate box selection stage. results traditional can only select a relatively optimal maximum box. If is not accurate enough, this type will be able do adjust it. To solve problem, an multiple fusion proposed. retain and delete non‐maximum box, but also position again. Thereby more obtained. In order verify generalization ability method, combined with two frameworks: faster R‐CNN model YOLOv3 model. these experiments prove that proposed achieve higher accuracy complete task effectively.