Research map / Point clouds / Segmentation and detection
All in One: Visual-Description-Guided Unified Point Cloud Segmentation
Zongyan Han, Mohamed El Amine Boudjoghra, Jiahua Dong, Jinhong Wang, Rao Muhammad Anwer
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 LLM description queries, image queries, SVC loss and spatial enhancement to OneFormer3D | OneFormer3D · S3DIS Area-5 | PQ: 62.2 -> 66.3 | Improved |
| add Spatial Enhancement Module (ablation) | VDG-Uni3DSeg · S3DIS Area-5 | mAP: 58.1 -> 60.1 | Improved |
| add description queries (ablation) | VDG-Uni3DSeg · S3DIS Area-5 | mIoU: 70.2 -> 71.5 | Improved |
| add image queries (ablation) | VDG-Uni3DSeg · S3DIS Area-5 | mIoU: 70.4 -> 71.5 | Improved |
| add Semantic-Visual Contrastive loss (ablation) | VDG-Uni3DSeg · S3DIS Area-5 | PQ: 62.2 -> 66.3 | Improved |
| use CLIP-selected top-5 internet images instead of random images (ablation) | VDG-Uni3DSeg · S3DIS Area-5 | PQ: 62.0 -> 66.3 | Improved |
| increase sparse-attention sample size from 16 to 128 (ablation) | Spatial Enhancement Module · ScanNet validation | PQ: 70.1 -> 71.5 | Improved |
| apply VDG-Uni3DSeg instead of OneFormer3D on ScanNet hidden test | OneFormer3D · ScanNet hidden test | mAP25: 89.6 -> 88.0 | Hurt |
| apply VDG-Uni3DSeg instead of OneFormer3D on ScanNet200 | OneFormer3D · ScanNet200 validation | mAP50: 40.0 -> 40.0 | No clear effect |
| apply VDG-Uni3DSeg instead of OneFormer3D on S3DIS 6-fold | OneFormer3D · S3DIS 6-fold cross-validation | mIoU: 72.8 -> 73.2 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- OneFormer3D — OneFormer3D: One Transformer for Unified Point Cloud Segmentation (CVPR 2024) · better on 11 of 12 shared results
- PointNeXt-XL — PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies (NeurIPS 2022) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Unified Framework for 3D Scene Understanding (NeurIPS 2024 · similarity 0.90)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.89)
- LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds (CVPR 2025 · similarity 0.89)
- 3D-AVS: LiDAR-based 3D Auto-Vocabulary Segmentation (CVPR 2025 · similarity 0.89)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.89)
- Multimodality Helps Few-shot 3D Point Cloud Semantic Segmentation (ICLR 2025 · similarity 0.89)
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.
