Research map / 3D vision / 3D generation
Emergent Outlier View Rejection in Visual Geometry Grounded Transformers
Jisang Han, Sunghwan Hong, Jaewoo Jung, Wooseok Jang, Honggyu An, Qianqian Wang, Seungryong Kim, Chen Feng
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 |
|---|---|---|---|
| filter distractor views via final-layer feature similarity before reconstruction | VGGT · Phototourism | ATE: 0.3504 -> 0.2650 | Improved |
| filter distractor views via final-layer feature similarity | VGGT · On-the-Go | ATE: 0.1210 -> 0.0583 | Improved |
| filter distractor views via final-layer feature similarity | VGGT · ETH3D | ATE: 0.9306 -> 0.7633 | Improved |
| filter distractors with attention-based score | VGGT · ETH3D multi-view depth | AbsRel: 0.0419 -> 0.0324 | Improved |
| use DINOv3 global descriptors for view pre-filtering | VGGT · Phototourism | success rate of distractor rejection: 0.000 average success | Hurt |
| use RobustVGGT-F instead of oracle-tuned MegaLoc+VGGT | VGGT · ETH3D | distractor rejection success: 0.887 -> 0.985 | Improved |
| use RobustVGGT instead of MASt3R-SfM | VGGT · RobustNeRF | ATE: 0.1177 vs 0.1379 | Mixed |
| lower attention threshold from 0.05 to 0.02 (ablation) | RobustVGGT-A · Phototourism | ATE: 0.2818 -> 0.3418 | Hurt |
| lower feature threshold from 0.65 to 0.35 (ablation) | RobustVGGT-F · Phototourism | ATE: 0.2650 -> 0.2777 | Hurt |
| increase distractors from Small to Large | VGGT without filtering · Phototourism | ATE: 0.3068 -> 0.3833 | Hurt |
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 6 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
- Robust Neural Rendering in the Wild with Asymmetric Dual 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.86)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.86)
- View Transformation Robustness for Multi-View 3D Object Reconstruction with Reconstruction Error-Guided View Selection (AAAI 2025 · similarity 0.86)
- RegimeVGGT: Layer-Wise Spatially Preserving Redundancy Removal for Visual Geometry Grounded Transformer (NeurIPS 2026 · similarity 0.85)
- Outlier-aware Slicing for Post-Training Quantization in Vision Transformer (ICML 2024 · similarity 0.85)
- The More You See in 2D the More You Perceive in 3D (CVPR 2024 · 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.
