Research map / Point clouds / Segmentation and detection
Pseudo-Embedding for Generalized Few-Shot Point Cloud Segmentation
Chih-Jung Tsai, Hwann-Tzong Chen, Tyng-Luh Liu
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 |
|---|---|---|---|
| generate pseudo-classes from background points via foundation model text grounding | generalized few-shot 3D segmentation · S3DIS 1-shot | HM: 41.92 -> 51.34 | Improved |
| generate pseudo-classes from background points | generalized few-shot 3D segmentation · S3DIS 5-shot | mIoU-N: 43.26 -> 50.23 | Improved |
| generate pseudo-classes from background points | generalized few-shot 3D segmentation · ScanNet 1-shot | mIoU-N: 14.78 -> 15.57 | Improved |
| generate pseudo-classes from background points | generalized few-shot 3D segmentation · ScanNet 5-shot | mIoU-N: 18.58 -> 23.34 | Improved |
| pseudo-classes vs baselines on base classes ceiling, floor, door | S3DIS 5-shot · S3DIS 5-shot base classes | IoU: ceiling 92.62 vs 93.17 (CAPL) | No clear effect |
| pseudo-classes vs baselines on clutter class | S3DIS 5-shot · S3DIS 5-shot base classes | IoU (clutter): 60.66 -> 57.82 | Hurt |
| pseudo-classes on novel column class | S3DIS 5-shot · S3DIS 5-shot novel classes | IoU (column): 32.16 -> 27.34 | Hurt |
| use pseudo-classes rather than single background class | support set variance · S3DIS 5-shot | HM std: 2.75 -> 2.00 | Improved |
| treat each point as a small sphere when projecting to images (ablation) | pseudo-class generation on ScanNet · ScanNet novel points | mAcc: 28.02 -> 48.16 | Improved |
Closest papers by meaning
- Generalized Few-shot 3D Point Cloud Segmentation with Vision-Language Model (CVPR 2025 · similarity 0.93)
- Rethinking Few-shot 3D Point Cloud Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (CVPR 2025 · similarity 0.88)
- Generated and Pseudo Content guided Prototype Refinement for Few-shot Point Cloud Segmentation (NeurIPS 2024 · similarity 0.88)
- Generalized Zero-Shot Learning for Point Cloud Segmentation with Evidence-Based Dynamic Calibration (AAAI 2025 · similarity 0.88)
- A Surprisingly Simple Approach to Generalized Few-Shot Semantic Segmentation (NeurIPS 2024 · 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.
