Research map / 3D vision / Depth and stereo
Diversity-Aware View Partitioning for Scalable VGGT
Jinsoo Park, Donggyu Choi, Ahyun Seo, Minsu Cho, Jeany Son
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 |
|---|---|---|---|
| use diversity-aware partitioning instead of random partitioning (ablation) | VGGT chunking · NRGBD pose estimation | AUC@3: 0.6819 -> 0.7844 | Improved |
| use diversity-aware partitioning instead of sequential partitioning (ablation) | VGGT chunking · NRGBD pose estimation | AUC@3: 0.3524 -> 0.7844 | Improved |
| use diversity-aware partitioning instead of K-means (ablation) | VGGT chunking · NRGBD pose estimation | AUC@3: 0.3557 -> 0.7844 | Improved |
| add visual dissimilarity to random initial partition (ablation) | partitioning · NRGBD pose estimation | AUC@3: 68.19 -> 78.21 | Improved |
| add spatial dispersion to random initial partition (ablation) | partitioning · NRGBD pose estimation | AUC@3: 68.19 -> 77.87 | Improved |
| combine visual dissimilarity and spatial dispersion instead of visual alone (ablation) | partitioning · NRGBD pose estimation | AUC@3: 78.21 -> 78.44 | Improved |
| use translation-only soft pose cue instead of translation plus rotation (ablation) | soft pose propagation · 7Scenes pose estimation | AUC@3: 0.2361 -> 0.2475 | Improved |
| use dissimilarity instead of raw similarity as score (ablation) | partitioning · NRGBD 3D reconstruction | Chamfer distance: 0.243 -> 0.029 | Improved |
| run 5 local search iterations instead of none (ablation) | partitioning · 7Scenes pose estimation | AUC@3: 0.2248 -> 0.2486 | Improved |
| increase chunk size from 25 to 100 at 1000 frames (ablation) | VGGT* + Ours · 7Scenes pose estimation | AUC@3: 0.2399 -> 0.2478 | Improved |
| use more anchor frames (20 instead of 1) (ablation) | chunk alignment · TUM-RGBD | ATE: 2.07 -> 1.97 | No clear effect |
| replace DINOv2 features with SALAD or MegaLoc descriptors (ablation) | dissimilarity source · NRGBD pose estimation | AUC@3: 0.7844 -> 0.7799 | No clear effect |
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.
- VGGT* — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 8 of 10 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
- VGGT: Visual Geometry Grounded Transformer (CVPR 2025 · similarity 0.89)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.88)
- RegimeVGGT: Layer-Wise Spatially Preserving Redundancy Removal for Visual Geometry Grounded Transformer (NeurIPS 2026 · similarity 0.88)
- LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token Merging (CVPR 2026 · similarity 0.87)
- FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry (NeurIPS 2026 · similarity 0.86)
- VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold (NeurIPS 2025 · 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.
