An advanced adaptive detector for oriented objects in remote sensing imagery.

ABSTRACT: YOLOv8 excels in generic horizontal bounding box detection but faces significant challenges when detecting oriented objects in remote sensing imagery. Targets in remote sensing images often exhibit large-scale variations, irregular shapes, and arbitrary orientations, which pose substantial difficulties for oriented detection. Additionally, inherent issues such as inconsistency in loss measurement and angle discontinuity further complicate oriented detection. Existing methods typically address only singular aspects of these challenges. In this paper, we propose an advanced adaptive detector based on YOLOv8, named AdaR-YOLOv8, which comprises three innovative strategies.For feature extraction, we designed the C2f-DCN module based on deformable convolutions, allowing the convolutions to dynam

SUBMITTER: Li Q 

PROVIDER: S-EPMC12484878 | biostudies-literature | 2025 Sep

REPOSITORIES: biostudies-literature