Rotation-Invariant Feature Enhancement with Dual-Aspect Loss for Arbitrary-Oriented Object Detection in Remote Sensing
Published in May 7, 2025
This paper proposes RFE-FCOS, which focuses on improving object detection in remote sensing imagery by incorporating multi-angle rotation-invariant learning. The method significantly enhances detection performance, particularly for arbitrarily oriented objects, achieving robust results on the DIOR-R and HRSC2016 benchmarks.