Oriented ship detector improves accuracy with strip convolution and exclusion loss
Strip Convolution and Direction-Aware Exclusion Loss for Oriented Ship Detection
Computer Vision and Pattern Recognition
Summary
Detecting ships that are oriented at different angles in very detailed satellite images is tricky because ships are long and there can be many crowds of them together. The authors designed a new ship detection system that uses special filters shaped like strips to better recognize the long shape of ship hulls. They also created a new way to reduce repeated detections by considering the ship's type, direction, and confidence score. Their approach improves accuracy on standard ship image datasets with relatively few model parameters.
What this means in practice
- •For maritime surveillance teams: Automatically detect ships with precise orientation in crowded port images to improve monitoring and tracking efficiency.
- •For aerial mapping service providers: Enhance vessel detection accuracy in high resolution aerial imagery for mapping and cataloging maritime assets.
Authors
Bin Chen, Yuanyuan Liu, Peng Yang, Chao Lu
Abstract
Oriented ship detection in very high resolution (VHR) remote sensing imagery remains challenging due to elongated hull geometry and dense target distributions in complex port scenes. Existing methods typically address geometric representation and duplicate suppression separately. To jointly tackle these issues, we propose an oriented ship detector with two complementary components. The C3k2_Strip module employs orthogonal strip convolutions to better capture elongated hull structures, while the Class-Aware Direction-Aware Exclusion Loss (CA-DAEL) suppresses redundant predictions using class, direction, and confidence cues. Experiments on HRSC2016 and DIOR-R achieve 78.45% and 53.71% mAP50:95, respectively, with only 2.91M parameters. On HRSC2016, the proposed method improves mAP50:95 by 6.32 percentage points over the YOLOv11-OBB baseline, demonstrating its effectiveness for accurate oriented ship detection.