Research map / Detection and segmentation / Open-set and few-shot detection
OpenVoxel: Training-Free Grouping and Captioning Voxels for Open-Vocabulary 3D Scene Understanding
Sheng-Yu Huang, Jaesung Choe, Yu-Chiang Frank Wang, Cheng Sun
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 mask merging of overlapping small masks during voxel grouping (ablation) | OpenVoxel · Ref-LeRF | mIoU: 24.3 -> 28.0 | Improved |
| add canonical captioning of each group in the scene map (ablation) | OpenVoxel · Ref-LeRF | mIoU: 28.0 -> 36.4 | Improved |
| add canonicalization of the input query into the caption template (ablation) | OpenVoxel · Ref-LeRF | mIoU: 36.4 -> 42.4 | Improved |
| replace learned sentence embeddings (ReferSplat) with training-free text-to-text retrieval over captions | RES · Ref-LeRF | avg mIoU: 29.2 -> 42.4 | Improved |
| compared to reproduced ReferSplat | RES · Ref-LeRF | avg mIoU: 24.5 -> 42.4 | Improved |
| use shared hyperparameters from the kitchen scene across all scenes | OpenVoxel · Ref-LeRF | avg mIoU: 42.4 -> 41.0 | No clear effect |
| training-free grouping and captioning instead of gradient-based language field training | OpenVoxel · scene map construction | run time: >1 hr -> ~3 min | Improved |
| use OVS-oriented prompt with training-free text retrieval | OpenVoxel · LeRF-OVS | avg mIoU: 65.1 -> 66.2 | Improved |
| OpenVoxel vs ObjectGS on LeRF-Mask | OpenVoxel · LeRF-Mask | avg mIoU: 88.3 -> 89.7 | Improved |
| OpenVoxel vs best baseline on LeRF-OVS figurines | OpenVoxel · LeRF-OVS figurines | mIoU: 73.5 -> 60.7 | Hurt |
Closest papers by meaning
- OpenLex3D: A Tiered Benchmark for Open-Vocabulary 3D Scene Representations (NeurIPS 2025 · similarity 0.90)
- Mosaic3D: Foundation Dataset and Model for Open-Vocabulary 3D Segmentation (CVPR 2025 · similarity 0.88)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.88)
- Cross-Modal and Uncertainty-Aware Agglomeration for Open-Vocabulary 3D Scene Understanding (CVPR 2025 · similarity 0.87)
- OpenNeRF: Open Set 3D Neural Scene Segmentation with Pixel-Wise Features and Rendered Novel Views (ICLR 2024 · similarity 0.87)
- COS3D: Collaborative Open-Vocabulary 3D Segmentation (NeurIPS 2025 · 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.
