Research map / 3D vision / 3D generation
Open Vocabulary 3D Scene Understanding via Geometry Guided Self-Distillation
Pengfei Wang, Yuxi Wang, Shuai Li, Zhaoxiang Zhang, Zhen Lei, Lei Zhang
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 superpoint-based geometry guided distillation loss (ablation) | pixel-point distillation · ScanNet | mIoU: 52.9 -> 53.5 | Improved |
| add self-distillation with EMA teacher (ablation) | geometry guided distillation · ScanNet | mIoU: 53.5 -> 56.1 | Improved |
| add superpoint voting to self-distillation pseudo-labels (ablation) | self-distillation · ScanNet | mIoU: 56.1 -> 56.5 | Improved |
| use EMA model instead of fixed 3D model for self-distillation targets (ablation) | GGSD · ScanNet | mIoU: 55.0 -> 56.1 | Improved |
| use fixed 2D model predictions as self-distillation signal (ablation) | GGSD · ScanNet | mIoU: 54.0 -> 51.9 | Hurt |
| refine 2D features with SAM masks instead of 3D geometry (ablation) | pixel-point distillation · ScanNet | mIoU: 52.9 -> 53.0 | No clear effect |
| distill 3D student from 2D teacher (pixel-point distillation) (ablation) | 2D fusion projection · ScanNet | mIoU: 50.0 -> 52.9 | Improved |
| apply GGSD to OpenScene baseline | OpenScene (pure 3D) · nuScenes | mIoU: 42.9 -> 46.1 | Improved |
| apply GGSD to tail categories | OpenScene (pure 3D) · ScanNet extended labels (Tail) | mIoU: 14.9 -> 16.0 | Mixed |
| zero-shot transfer to Matterport3D with more categories | GGSD · Matterport160 | mIoU: 6.0 -> 6.3 | Improved |
Closest papers by meaning
- Cross-Modal and Uncertainty-Aware Agglomeration for Open-Vocabulary 3D Scene Understanding (CVPR 2025 · similarity 0.89)
- OpenLex3D: A Tiered Benchmark for Open-Vocabulary 3D Scene Representations (NeurIPS 2025 · similarity 0.89)
- OpenGaussian: Towards Point-Level 3D Gaussian-based Open Vocabulary Understanding (NeurIPS 2024 · similarity 0.88)
- Lexicon3D: Probing Visual Foundation Models for Complex 3D Scene Understanding (NeurIPS 2024 · similarity 0.88)
- PanoGS: Gaussian-based Panoptic Segmentation for 3D Open Vocabulary Scene Understanding (CVPR 2025 · similarity 0.87)
- Open3DSG: Open-Vocabulary 3D Scene Graphs from Point Clouds with Queryable Objects and Open-Set Relationships (CVPR 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.
