Research map / Detection and segmentation / CNN features
RDD: Improving Structure from Motion with a Robust Feature Detector and Descriptor
G Chen, Tianwen Fu, Haiwei Chen, Wenbin Teng, Hanyuan Xiao, Yajie Zhao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add deformable-transformer descriptor branch (RDD) instead of best prior detector (DeDoDe-V2-G), MNN matching | RDD · MegaDepth-View | AUC@5deg: 33.1 -> 38.3 | Improved |
| use RDD instead of DeDoDe-V2-G on Air-to-Ground pairs | sparse MNN matching · Air-to-Ground benchmark | AUC@5deg: 31.5 -> 41.4 | Improved |
| use larger patch size s=8 (ablation) | RDD · MegaDepth-1500 | AUC@5deg: 48.2 -> 44.6 | Hurt |
| use fewer sampled points per query (N=4) in deformable attention (ablation) | RDD · MegaDepth-1500 | AUC@5deg: 48.2 -> 46.5 | Hurt |
| add the dedicated keypoint branch (ablation) | RDD · MegaDepth-1500 | AUC@5deg: 44.1 -> 48.2 | Improved |
| train the two branches separately instead of jointly (ablation) | RDD · MegaDepth-1500 | AUC@5deg: 42.8 -> 48.2 | Improved |
| train the two branches separately instead of jointly (ablation) | RDD* · MegaDepth-1500 | AUC@5deg: 46.9 -> 51.3 | Improved |
| add epipolar match refinement to semi-dense matching (ablation) | RDD* · MegaDepth-1500 | AUC@5deg: 41.3 -> 51.3 | Improved |
| add Air-to-Ground data to training (ablation) | RDD · MegaDepth-1500 | AUC@5deg: 47.4 -> 48.2 | Improved |
| add Air-to-Ground data to training (ablation) | RDD* · MegaDepth-1500 | AUC@5deg: 50.4 -> 51.3 | Improved |
| use larger patch size s=8 (ablation) | RDD* · MegaDepth-1500 | AUC@5deg: 51.3 -> 49.4 | Hurt |
| use LightGlue matcher instead of MNN | RDD · MegaDepth-View | AUC@5deg: 38.3 -> 54.2 | Improved |
Closest papers by meaning
- RDD: Robust Feature Detector and Descriptor using Deformable Transformer (CVPR 2025 · similarity 0.98)
- Enhancing Deformable Local Features by Jointly Learning To Detect and Describe Keypoints (CVPR 2023 · similarity 0.86)
- KeypointDETR: An End-to-End 3D Keypoint Detector (ECCV 2024 · similarity 0.85)
- The Drunkard’s Odometry: Estimating Camera Motion in Deforming Scenes (NeurIPS 2023 · similarity 0.85)
- GEOBIT: A Geodesic-Based Binary Descriptor Invariant to Non-Rigid Deformations for RGB-D Images (ICCV 2019 · similarity 0.85)
- DV-Matcher: Deformation-based Non-rigid Point Cloud Matching Guided by Pre-trained Visual Features (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
