Research map / Vision-language and multimodal / CLIP and zero-shot
UniPart: Towards Zero-shot Language-Grounded 3D Part Segmentation for Embodied Interaction
Xinqiang Yu, Zekun Qi, Jiawei He, Wenyao Zhang, Xuchuan Chen, Guocai Yao, Li Yi, Zhaoxiang Zhang, HE WANG
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 |
|---|---|---|---|
| replace FIND3D architecture with UniPart under same FIND3D training data (ablation) | UniPart · Objaverse-General zero-shot | mIoU: 30.75 -> 35.27 | Improved |
| train FIND3D model on LangPart-1M instead of FIND3D data (ablation) | FIND3D · Objaverse-General zero-shot | mIoU: 30.75 -> 38.42 | Improved |
| train UniPart on LangPart-1M instead of FIND3D data (ablation) | UniPart · Objaverse-General zero-shot | mIoU: 35.27 -> 47.25 | Improved |
| add CLIP-alignment pretraining on partial point clouds (ablation) | UniPart · LangPart-1K single-view | mIoU: 26.34 -> 33.56 | Improved |
| add CLIP-alignment pretraining (ablation) | UniPart · LangPart-1K whole-object | mIoU: 23.39 -> 27.17 | Improved |
| additive text injection instead of concatenation (ablation) | UniPart · LangPart-1K | mIoU: 30.31 -> 33.56 | Improved |
| additive text injection instead of cross-attention (ablation) | UniPart · LangPart-1K | mIoU: 29.22 -> 33.56 | Improved |
| additive text injection instead of multiplication (ablation) | UniPart · LangPart-1K | mIoU: 28.13 -> 33.56 | Improved |
| UniPart vs PartSLIP++ with short prompt on PartNet-E | UniPart · PartNet-E canonical | mIoU: 32.71 (PartSLIP++) vs 22.90 (UniPart) | Hurt |
Closest papers by meaning
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.86)
- Articulate your NeRF: Unsupervised articulated object modeling via conditional view synthesis (NeurIPS 2024 · similarity 0.86)
- InteractAnything: Zero-shot Human Object Interaction Synthesis via LLM Feedback and Object Affordance Parsing (CVPR 2025 · similarity 0.86)
- Generative 3D Part Assembly via Part-Whole-Hierarchy Message Passing (CVPR 2024 · similarity 0.86)
- Part-X-MLLM: Part-aware 3D Multimodal Large Language Model (ICLR 2026 · similarity 0.86)
- UniPose: A Unified Multimodal Framework for Human Pose Comprehension, Generation and Editing (CVPR 2025 · similarity 0.85)
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.
