Research map / 3D vision / 3D generation
Towards Geometry-Grounded Dense Semantic Matching with VGGT Priors
Songlin Yang, Tianyi Wei, Yushi Lan, Zeqi Xiao, Anyi Rao, Xingang Pan
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 sparse keypoint supervision to untuned VGGT (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 9.0 -> 79.4 | Improved |
| add sparse supervision (ablation) | semantic matching model on VGGT · Synthetic dense test | dense MSE: 0.49 -> 0.18 | Hurt |
| add synthetic data (ablation) | semantic matching model on VGGT · Synthetic dense test | dense MSE: 0.18 -> 0.11 | Improved |
| add synthetic data (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 79.4 -> 79.3 | No clear effect |
| add smoothness loss (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 79.3 -> 79.1 | No clear effect |
| add reconstruction loss (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 79.1 -> 80.5 | Improved |
| add DINO feature matching loss instead of VGG loss (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 81.8 -> 84.2 | Improved |
| initialise from pretrained VGGT instead of random (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 69.2 -> 84.2 | Improved |
| use VGGT instead of DUSt3R (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 58.7 -> 84.2 | Improved |
| share 4 frozen early blocks rather than 18 (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 59.3 -> 84.2 | Improved |
| share 4 frozen early blocks rather than 12 (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 65.0 -> 84.2 | Improved |
| share 4 blocks rather than 6 (ablation) | semantic matching model on VGGT · SPair-71k | PCK@0.1: 77.2 -> 84.2 | Improved |
Closest papers by meaning
- Bridging Viewpoint Gaps: Geometric Reasoning Boosts Semantic Correspondence (CVPR 2025 · similarity 0.87)
- Pixel-level Semantic Correspondence through Layout-aware Representation Learning and Multi-scale Matching Integration (CVPR 2024 · similarity 0.87)
- Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning (CVPR 2026 · similarity 0.86)
- LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token Merging (CVPR 2026 · similarity 0.86)
- Jamais Vu: Exposing the Generalization Gap in Supervised Semantic Correspondence (NeurIPS 2025 · similarity 0.86)
- G^3-LQ: Marrying Hyperbolic Alignment with Explicit Semantic-Geometric Modeling for 3D Visual Grounding (CVPR 2024 · similarity 0.86)
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.
