CoReMatch: Recurrent Continuous Coordinate Refinement for Real-Time Robotic Feature Matching
CoReMatch: Recurrent Continuous Coordinate Refinement for Real-Time Robotic Feature Matching
This paper presents CoReMatch, a recurrent framework for real-time robotic feature matching that refines correspondences directly in continuous coordinate space. By using a Deep Aggregated Transformer at the coarse stage and a lightweight regression head to sharpen localization, the method outperforms Efficient LoFTR in accuracy while running more than 10x faster on embedded edge hardware, demonstrating an efficient and scalable approach for robotic perception tasks like visual localization, mapping, SLAM, and autonomous navigation.
The authors introduce CoReMatch, a recurrent framework that refines feature correspondences in continuous coordinate space to achieve higher accuracy and speed on embedded hardware compared to existing semi-dense matchers.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.03932v1 Announce Type: new Abstract: Feature matching underpins many robotic perception tasks, including visual localization, mapping, SLAM, and autonomous navigation. Existing semi-dense matchers predict confidence on fixed grids, limiting correspondences to predefined locations and requiring costly high-resolution grids for precise localization. We present CoReMatch (Coordinate Refinement Match), a recur
Open source S2
rent framework that refines correspondences directly in continuous coordinate space rather than predicting discrete confidence maps. CoReMatch progressively improves correspondence estimates by propagating information across iterations. At the coarse stage, a Deep Aggregated Transformer captures global context and iteratively updates correspondence coordinates, while a lightweight regression head further sharpens the
Open source S3
ir localization. Experiments on indoor and outdoor benchmarks show that CoReMatch strikes a strong balance between accuracy and efficiency, outperforming Efficient LoFTR in accuracy while running more than 10x faster on embedded edge hardware. These results demonstrate that recurrent coordinate refinement offers an efficient and scalable approach to real-time feature matching for robotic perception.
Open source S4
CoReMatch: Recurrent Continuous Coordinate Refinement for Real-Time Robotic Feature Matching
Open source S1
Source：arXiv Robotics — research abstracts · arxiv.org