Research map / Architectures and efficiency / Attention and transformers
Adaptive Spot-Guided Transformer for Consistent Local Feature Matching
Jiahuan Yu, Jiahao Chang, Jianfeng He, Tianzhu Zhang, Jiyang Yu, Feng Wu
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 multi-level cross attention at 1/32 resolution (ablation) | ASTR baseline · MegaDepth | pose estimation AUC@5°: 45.6 -> 46.7 | Improved |
| add spot-guided attention at 1/8 resolution in place of linear attention (ablation) | ASTR with multi-level cross attention · MegaDepth | pose estimation AUC@5°: 46.7 -> 47.7 | Improved |
| add adaptive scaling (ablation) | ASTR with multi-level and spot-guided attention · MegaDepth | pose estimation AUC@5°: 47.7 -> 48.3 | Improved |
| vary k from 1 to 6 with l=5 (ablation) | ASTR spot-guided attention · MegaDepth | pose estimation AUC@5°: 46.0; 47.5; 47.3; 47.7; 47.1; 46.9 | Mixed |
| vary l from 3 to 9 with k=4 (ablation) | ASTR spot-guided attention · MegaDepth | pose estimation AUC@5°: 46.7; 47.7; 47.2; 43.0 | Mixed |
| use ASTR instead of PDC-Net+ | HPatches homography estimation · HPatches | homography estimation AUC: +4.4% at 3px; +3.5% at 5px; +2.5% at 10px | Improved |
| use ASTR instead of ASpanFormer | MegaDepth relative pose estimation · MegaDepth | pose estimation AUC: +3.1% at 5°; +1.6% at 10° | Improved |
| use ASTR instead of Quadtree | ScanNet relative pose estimation (models not trained on ScanNet) · ScanNet | pose estimation AUC: 19.0 -> 19.4; 37.3 -> 37.6; 53.5 -> 54.4 | Improved |
| use ASTR instead of ASpanFormer | InLoc visual localization · InLoc DUC1 | registered images (%): 51.5 -> 53.0; 73.7 -> 73.7; 86.4 -> 87.4 | Mixed |
| use ASTR instead of ASpanFormer | Aachen visual localization · Aachen Day and Night | registered images (%): 89.4 -> 89.9; 95.6 -> 95.6; 99.0 -> 99.2; 77.5 -> 76.4; 91.6 -> 92.1; 99.0 -> 99.5 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ASpanFormer — ASpanFormer: Detector-Free Image Matching with Adaptive Span Transformer (ECCV 2022) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Consistency-Aware Spot-Guided Transformer for Versatile and Hierarchical Point Cloud Registration (NeurIPS 2024 · similarity 0.84)
- Consistency-Aware Spot-Guided Transformer for Accurate and Versatile Point Cloud Registration (TPAMI 2026 · similarity 0.84)
- ETO:Efficient Transformer-based Local Feature Matching by Organizing Multiple Homography Hypotheses (NeurIPS 2024 · similarity 0.83)
- Semi-Supervised Wide-Angle Portraits Correction by Multi-Scale Transformer (CVPR 2022 · similarity 0.83)
- Dynamic Cues-Assisted Transformer for Robust Point Cloud Registration (CVPR 2024 · similarity 0.81)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.80)
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.
