Research map / 3D vision / 3D generation
Multi-View Foundation Models
Leo Segre, Or Hirschorn, Shai Avidan
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 LoRA training with correspondence loss (ablation) | DINOv2 · ScanNet++ correspondence | localization error: 0.1029 -> 0.0347 | Improved |
| add multi-view adapter (ablation) | DINOv2 + LoRA · ScanNet++ correspondence | localization error: 0.0347 -> 0.0262 | Improved |
| add multi-view adapter (ablation) | DINOv2 + LoRA · base similarity | cosine similarity: 0.9117 -> 0.9360 | Improved |
| add Plucker ray embedding (ablation) | multi-view adapter · ScanNet++ correspondence | localization error: 0.0262 -> 0.0247 | Improved |
| add regularization to base features (ablation) | full method · base similarity | cosine similarity: 0.1054 -> 0.9376 | Improved |
| add regularization to base features (ablation) | full method · ScanNet++ correspondence | localization error: 0.0204 -> 0.0247 | Hurt |
| remove camera poses (pose-free variant) (ablation) | Ours · DINOv2 similarity generalization | cosine similarity: 0.8003 -> 0.8099 | No clear effect |
| multi-view lifting vs FiT3D | DINOv2 · NAVI normals | RMSE: 51.00 -> 32.17 | Improved |
| multi-view lifting vs SnD | DINOv2 · ImageNet zero-shot KNN | top-1 accuracy: 56.31% -> 58.02% | Improved |
| lifting method applied to CLIP | CLIP · ScanNet++ correspondence | localization error: 0.1505 -> 0.0301 | Improved |
| lifting method applied to SAM | SAM · generalization correspondence | localization error: 0.1833 -> 0.0752 | Improved |
| add meaningful context views from 2 to 10 | Ours · ScanNet++ correspondence | localization error: 0.033 -> 0.029 | Improved |
Closest papers by meaning
- 3D-LFM: Lifting Foundation Model (CVPR 2024 · similarity 0.88)
- Probing the 3D Awareness of Visual Foundation Models (CVPR 2024 · similarity 0.86)
- All-in-One: Transferring Vision Foundation Models into Stereo Matching (AAAI 2025 · similarity 0.84)
- Interfacing Foundation Models' Embeddings (NeurIPS 2024 · similarity 0.84)
- Lexicon3D: Probing Visual Foundation Models for Complex 3D Scene Understanding (NeurIPS 2024 · similarity 0.84)
- Feat2GS: Probing Visual Foundation Models with Gaussian Splatting (CVPR 2025 · similarity 0.84)
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.
