Learning to zoom: Exploiting mixed-scale contextual information for object detection

Open Webpage

Boying Wang, Ruyi Ji, Libo Zhang, Yanjun Wu, Jing Liu

Published: 01 Jan 2025, Last Modified: 12 Nov 2025Expert Syst. Appl. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A novel Mixed-Scale Network(MSNet) is proposed to form an expressive representation.•Global feature aggregation module aggregates the features from mixed-scale images.•Global feature enhancement module alleviates the problem of information loss.•Local feature aggregation module further locally refine the instance feature.•Extensive experiments show that MSNet gains consistent performance.