Research map / Point clouds / Segmentation and detection
PartSLIP: Low-Shot Part Segmentation for 3D Point Clouds via Pretrained Image-Language Models
Minghua Liu, Yinhao Zhu, Hong Cai, Shizhong Han, Zhan Ling, Fatih Porikli, Hao Su
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 |
|---|---|---|---|
| Use PartSLIP few-shot prompt tuning and 3D lifting | PartNetE semantic segmentation · PartNetE (45 categories) | category mIoU (%): 44.3 -> 59.4 | Improved |
| Use PartSLIP few-shot prompt tuning and 3D lifting | PartNetE instance segmentation · PartNetE (45 categories) | category mAP50 (%): 31.9 -> 44.8 | Improved |
| Add 3D bounding-box voting and grouping (ablation) | GLIP zero-shot predictions · PartNetE Chair | category mIoU (%): 50.6 -> 60.7 | Improved |
| Add 3D bounding-box voting and grouping (ablation) | GLIP zero-shot predictions · PartNetE Kettle | category mIoU (%): 7.5 -> 20.8 | Improved |
| Add 3D bounding-box voting and grouping (ablation) | GLIP zero-shot predictions · PartNetE Suitcase | category mIoU (%): 21.1 -> 40.2 | Improved |
| Add 8-shot prompt tuning (ablation) | GLIP with 3D voting/grouping · PartNetE Kettle | category mIoU (%): 20.8 -> 72.7 | Improved |
| Add multi-view feature aggregation (ablation) | GLIP with 3D voting/grouping · PartNetE Suitcase | category mIoU (%): 40.2 -> 49.1 | Improved |
| Fuse visual features early, before vision-language fusion (ablation) | GLIP multi-view aggregation · PartNetE Suitcase | 2D detection mAP50: 65.6 -> 68.9 | Improved |
| Fuse visual features late, before the detection head (ablation) | GLIP multi-view aggregation · PartNetE Suitcase | 2D detection mAP50: 65.6 -> 47.3 | Hurt |
| Use partial point cloud input (ablation) | PartSLIP · PartNetE Chair | category mIoU (%): 85.3 -> 84.3 | No clear effect |
| Remove texture and use grayscale input (ablation) | PartSLIP · PartNetE Chair | category mIoU (%): 85.3 -> 84.0 | No clear effect |
| Make point cloud sparse at 128×128 resolution (ablation) | PartSLIP · PartNetE Chair | category mIoU (%): 85.3 -> 82.4 | Hurt |
Closest papers by meaning
- Generalized Few-shot 3D Point Cloud Segmentation with Vision-Language Model (CVPR 2025 · similarity 0.90)
- S$^2$AM3D: Scale-controllable Part Segmentation of 3D Point Clouds (CVPR 2026 · similarity 0.89)
- PartSAM: A Scalable Promptable Part Segmentation Model Trained on Native 3D Data (ICLR 2026 · similarity 0.89)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.88)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.88)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (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.
