Research map / 3D vision / 3D generation
MVGGT: Multimodal Visual Geometry Grounded Transformer for Multiview 3D Referring Expression Segmentation
Changli Wu, Haodong Wang, Jiayi Ji, Yutian Yao, Chunsai Du, Jihua Kang, Yanwei Fu, Liujuan Cao
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 PVSO (ablation) | MVGGT without multimodal branch · MVRefer overall | mIoU global: 26.9 -> 32.0 | Improved |
| add PVSO (ablation) | MVGGT without multimodal branch · MVRefer overall | mIoU view: 41.1 -> 47.5 | Improved |
| add the MVGGT multimodal branch with language guidance (ablation) | MVGGT with PVSO · MVRefer overall | mIoU global: 32.0 -> 39.9 | Improved |
| add the MVGGT multimodal branch with language guidance (ablation) | MVGGT with PVSO · MVRefer hard | mIoU global: 19.0 -> 24.4 | Improved |
| add no-target suppression to random sampling (ablation) | PVSO · MVRefer overall | mIoU global: 32.4 -> 36.7 | Improved |
| use hybrid sampling with no-target ratio 0.5 instead of 0.25 (ablation) | PVSO · MVRefer overall | mIoU global: 35.7 -> 39.9 | Improved |
| raise no-target ratio from 0.5 to 0.75 (ablation) | PVSO · MVRefer overall | mIoU global: 39.9 -> 25.7 | Hurt |
| use late fusion instead of early fusion of language (ablation) | MVGGT · MVRefer overall | mIoU global: 36.1 -> 39.9 | Improved |
| use late fusion instead of middle fusion of language (ablation) | MVGGT · MVRefer overall | mIoU global: 37.5 -> 39.9 | Improved |
| use 12 multimodal layers instead of 6 (ablation) | MVGGT · MVRefer overall | mIoU global: 37.3 -> 39.9 | Improved |
| expand multimodal branch from 12 to 16 layers (ablation) | MVGGT · MVRefer overall | mIoU global: 39.9 -> 25.4 | Hurt |
| use end-to-end MVGGT instead of two-stage reconstruct-then-segment | MVGGT vs two-stage · MVRefer overall | mIoU global: 18.5 -> 39.9 | Improved |
Closest papers by meaning
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.89)
- RG-SAN: Rule-Guided Spatial Awareness Network for End-to-End 3D Referring Expression Segmentation (NeurIPS 2024 · similarity 0.88)
- LESS: Label-Efficient and Single-Stage Referring 3D Segmentation (NeurIPS 2024 · similarity 0.88)
- VGGT: Visual Geometry Grounded Transformer (CVPR 2025 · similarity 0.88)
- RelationVGGT : Visual Geometry Transformers for 3D Spatial Relation Segmentation (NeurIPS 2026 · similarity 0.88)
- 3D-STMN: Dependency-Driven Superpoint-Text Matching Network for End-to-End 3D Referring Expression Segmentation (AAAI 2024 · similarity 0.87)
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.
