Research map / 3D vision / Human pose estimation
SGAD: Semantic and Geometric-aware Descriptor for Local Feature Matching
Xiangzeng Liu, Chi Wang, Guanglu Shi, Xiaodong Zhang, Qiguang Miao, Miao Fan
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 attention (self and cross) over area descriptors to DINOv2 features (ablation) | SGAD · ScanNet1500 area matching | AUC@0.2: 79.74 -> 92.52 | Improved |
| add geometric positional encoding (ablation) | SGAD · ScanNet1500 area matching | AUC@0.2: 92.52 -> 95.46 | Improved |
| add ranking loss to classification loss (ablation) | SGAD · ScanNet1500 area matching | AUC@0.2: 94.82 -> 95.46 | Improved |
| use dual-task classification and ranking loss instead of triplet loss (ablation) | SGAD · ScanNet1500 area matching | AUC@0.2: 91.81 -> 95.46 | Improved |
| replace MESA area matching with SGAD for LoFTR | LoFTR pipeline · MegaDepth runtime | runtime (s): 60.23 -> 0.82 | Improved |
| add SGAD to DKM versus MESA and DMESA (ScanNet 1296x968) | DKM · ScanNet1500 | pose AUC@5deg: 24.16 -> 31.63 | Improved |
| add SGAD to LoFTR at 1296x968 on ScanNet | LoFTR · ScanNet1500 | pose AUC@5deg: 8.91 -> 29.69 | Improved |
| add SGAD to ROMA | ROMA · ScanNet1500 | pose AUC@5deg: 31.51 -> 33.84 | Improved |
| add SGAD to LoFTR at low resolution on ScanNet 640x480 | LoFTR · ScanNet1500 (640x480) | pose AUC@5deg: 28.44 -> 28.69 | No clear effect |
| add SGAD to DKM on HPatches | DKM · HPatches | AUC@3px: 71.3 -> 72.1 | Improved |
| add SGAD to LoFTR on MegaDepth versus dense DKM | LoFTR · MegaDepth1500 (1200x1200) | pose AUC@5deg: 61.11 -> 65.98 | Improved |
Closest papers by meaning
- MESA: Effective Matching Redundancy Reduction by Semantic Area Segmentation (TPAMI 2025 · similarity 0.88)
- SAG-GNN: Semantic-Aware Guided GNN for Descriptor-Free 2D-3D Matching (CVPR 2026 · similarity 0.88)
- DGC-GNN: Leveraging Geometry and Color Cues for Visual Descriptor-Free 2D-3D Matching (CVPR 2024 · similarity 0.88)
- SemLight: Distilled Semantic–Geometric Fusion for Efficient Local Feature Matching (ECCV 2026 · similarity 0.88)
- GeoMix: Descriptor-Free Visual Localization via Global Context and Multi-Detector Training (ECCV 2026 · similarity 0.87)
- MESA: Matching Everything by Segmenting Anything (CVPR 2024 · similarity 0.87)
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.
