Research map / Detection and segmentation / Open-set and few-shot detection
DiSCO-3D : Discovering and Segmenting Sub-Concepts from Open-vocabulary Queries in NeRF
Doriand Petit, Steve Bourgeois, Vincent Gay-Bellile, Florian Chabot, Loïc Barthe
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 |
|---|---|---|---|
| perform USS and open-vocabulary guidance jointly instead of sequentially (vs USS->OVS) (ablation) | DiSCO-3D · Replica OV-SD LeRF | PQ: 4.76 -> 8.13 | Improved |
| perform jointly instead of OVS->USS (ablation) | DiSCO-3D · Replica OV-SD OpenNeRF | PQ: 5.47 -> 8.65 | Improved |
| add L_proto CLIP-prototype regularization (ablation) | DiSCO-3D · Replica OV-SD LeRF Hungarian (N=10) | PQ: 9.77 -> 10.19 | Improved |
| use ground-truth Hungarian matching instead of CLIP-prototype matching (ablation) | DiSCO-3D · Replica OV-SD LeRF | PQ: 8.13 -> 10.19 | Improved |
| replace USS module with K-Means in naive baselines (ablation) | naive OV-SD baselines · Replica OV-SD LeRF Hungarian | PQ: 6.94 -> 6.32 (USS->OVS vs K-Means->OVS) | Hurt |
| add DiSCO-3D on top of LeRF for OV-Seg classes | LeRF · Replica OV-Seg classes | mIoU: 8.79 -> 12.42 | Improved |
| add DiSCO-3D on top of OpenNeRF for OV-Seg concepts | OpenNeRF · Replica OV-Seg concepts | mAcc: 42.69 -> 58.17 | Improved |
| add DiSCO-3D on top of OpenNeRF for OV-Seg classes | OpenNeRF · Replica OV-Seg classes | mIoU: 21.60 -> 21.87 | No clear effect |
| add DiSCO-3D on top of OpenNeRF in mono-label paradigm | OpenNeRF · Replica OV-Seg mono-label | mAcc: 31.96 -> 30.19 | Hurt |
| add linear decay schedule of beta (ablation) | DiSCO-3D USS · Replica USS | mIoU: 26.52 -> 27.47 | Improved |
| weight prototype EMA update by both confidence and density (ablation) | DiSCO-3D USS · Replica USS | mIoU: 16.77 -> 27.47 | Improved |
| use full non-linear projector instead of two linear layers (ablation) | DiSCO-3D USS · Replica USS | mIoU: 27.12 -> 27.47 | No clear effect |
Closest papers by meaning
- COS3D: Collaborative Open-Vocabulary 3D Segmentation (NeurIPS 2025 · similarity 0.89)
- Masked Point-Entity Contrast for Open-Vocabulary 3D Scene Understanding (CVPR 2025 · similarity 0.88)
- Segment then Splat: Unified 3D Open-Vocabulary Segmentation via Gaussian Splatting (NeurIPS 2025 · similarity 0.88)
- PanoGS: Gaussian-based Panoptic Segmentation for 3D Open Vocabulary Scene Understanding (CVPR 2025 · similarity 0.88)
- A Unified Framework for 3D Scene Understanding (NeurIPS 2024 · similarity 0.88)
- OpenLex3D: A Tiered Benchmark for Open-Vocabulary 3D Scene Representations (NeurIPS 2025 · similarity 0.88)
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.
