RiO-DETR: DETR for Real-time Oriented Object Detection

2026-03-10

AI Summary

This work addresses three key challenges in adapting DETR to real-time oriented object detection: angular semantic dependency, periodicity-induced disruption of Euclidean optimization, and slow convergence due to an expanded search space. To overcome these issues, the authors propose a task-native design featuring content-driven angle estimation, rotation-corrected orthogonal attention, a decoupled shortest-path periodic loss, and an oriented dense one-to-one supervision strategy. This integrated approach enables efficient and stable angle learning with accelerated convergence. Extensive experiments on DOTA-1.0, DIOR-R, and FAIR1M-2.0 demonstrate that the proposed method achieves state-of-the-art performance, significantly improving the speed–accuracy trade-off for real-time oriented object detection.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

We present RiO-DETR: DETR for Real-time Oriented Object Detection, the first real-time oriented detection transformer to the best of our knowledge. Adapting DETR to oriented bounding boxes (OBBs) poses three challenges: semantics-dependent orientation, angle periodicity that breaks standard Euclidean refinement, and an enlarged search space that slows convergence. RiO-DETR resolves these issues with task-native designs while preserving real-time efficiency. First, we propose Content-Driven Angle Estimation by decoupling angle from positional queries, together with Rotation-Rectified Orthogonal Attention to capture complementary cues for reliable orientation. Second, Decoupled Periodic Refinement combines bounded coarse-to-fine updates with a Shortest-Path Periodic Loss for stable learning across angular seams. Third, Oriented Dense O2O injects angular diversity into dense supervision to speed up angle convergence at no extra cost. Extensive experiments on DOTA-1.0, DIOR-R, and FAIR-1M-2.0 demonstrate RiO-DETR establishes a new speed--accuracy trade-off for real-time oriented detection. Code will be made publicly available.

Problem

• oriented object detection
• real-time detection
• angle periodicity
• DETR adaptation
• rotated bounding boxes

Innovation

• oriented object detection
• DETR
• angle periodicity
• real-time detection
• rotation-rectified attention