Research map / Vision-language and multimodal / Cross-modal learning
SGPFeat: Semantic and Geometric Priors for Multi-modal Image Matching
Yuxin Deng, Botian Wang, Kaining Zhang, Hao Zhang, Jiayi Ma
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 semantic and geometric distillation losses to basic constraints (ablation) | SGPFeat · VIS-NIR (1024 keypoints) | SRR: 34.1 -> 81.1 | Improved |
| add semantic and geometric distillation losses (ablation) | SGPFeat · VIS-NIR (1024 keypoints) | Ravg (px): 6.10 -> 3.65 | Improved |
| add semantic and geometric distillation losses (ablation) | SGPFeat · VIS-NIR (1024 keypoints) | CLR (%): 10.6 -> 0 | Improved |
| add a single distillation loss to basic constraints (ablation) | SGPFeat · VIS-NIR (1024 keypoints) | SRR: 34.1 -> 45.5 | Improved |
| add two distillation losses (variant c) (ablation) | SGPFeat · VIS-NIR (1024 keypoints) | SRR: 34.1 -> 72.7 | Improved |
| add the third distillation loss on top of two (variant d to full) (ablation) | SGPFeat · VIS-NIR (1024 keypoints) | SRR: 71.2 -> 81.1 | Improved |
| replace ALIKE geometric teacher with RDD (ablation) | SGPFeat · VIS-IR | SRR: 87.2 -> 44.7 | Hurt |
| replace ALIKE geometric teacher with ALIKED (ablation) | SGPFeat · VIS-IR | Ravg (px): 3.94 -> 4.23 | Hurt |
| train on VIS-IR instead of VIS-NIR for zero-shot transfer (ablation) | SGPFeat · CT-MRI zero-shot | SRR: 58.3 -> 79.2 | Improved |
| train on VIS-IR instead of VIS-NIR for zero-shot transfer (ablation) | SGPFeat · EMA-OCTA zero-shot | SRR: 40.0 -> 50.0 | Improved |
| use DINO V2 instead of DepthAnything V2 as semantic teacher (ablation) | SGPFeat · VIS-NIR | matching accuracy | Hurt |
Closest papers by meaning
- Image-to-Image Matching via Foundation Models: A New Perspective for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.85)
- Self-Supervised Spatial Correspondence Across Modalities (CVPR 2025 · similarity 0.85)
- SSLFusion: Scale and Space Aligned Latent Fusion Model for Multimodal 3D Object Detection (AAAI 2025 · similarity 0.85)
- Topology-Aware Representation Alignment for Semi-Supervised Vision-Language Learning (NeurIPS 2026 · similarity 0.85)
- Noisy Correspondence Rectification via Asymmetric Similarity Learning (AAAI 2025 · similarity 0.85)
- OmniSegmentor: A Flexible Multi-Modal Learning Framework for Semantic Segmentation (NeurIPS 2025 · similarity 0.85)
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.
