Research map / Point clouds / Segmentation and detection
Dual-level Adaptive Self-Labeling for Novel Class Discovery in Point Cloud Segmentation
Ruijie Xu, CHUYU ZHANG, Hui Ren, Xuming He
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 equal-size constraint with imbalanced semi-relaxed OT self-labeling (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 31.8 -> 36.0 | Improved |
| add adaptive regularization of gamma (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 36.0 -> 44.2 | Improved |
| add region-level branch (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 44.2 -> 48.4 | Improved |
| adaptive gamma instead of best fixed gamma (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 36.0 -> 44.2 | Improved |
| use a very small fixed gamma (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 10.1 (gamma 0.01) vs 36.0 (gamma 0.5) | Hurt |
| use ground-truth class distribution as an equality constraint (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 32.5 vs adaptive 44.2 | No clear effect |
| use step decay or cosine annealing for gamma instead of KL-adaptive (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 34.6 and 36.1 vs 44.2 | Hurt |
| add DASL over NOPS on SemanticKITTI split 3 | DASL · SemanticKITTI split 3 | novel-class IoU: 12.4 -> 12.6 | No clear effect |
| use estimated number of novel classes (3 instead of 4) | DASL · SemanticPOSS split 0 | novel avg IoU: 31.95 -> 53.47 vs NOPS | Improved |
| increase DBSCAN-independent hyperparameter T to 30 (ablation) | DASL · SemanticPOSS split 0 | novel avg IoU: 44.21 -> 32.48 | Hurt |
Closest papers by meaning
- Novel Class Discovery for 3D Point Cloud Semantic Segmentation (CVPR 2023 · similarity 0.91)
- Self-Prediction for Joint Instance and Semantic Segmentation of Point Clouds (ECCV 2020 · similarity 0.90)
- Novel Class Discovery Meets Foundation Models for 3D Semantic Segmentation (IJCV 2024 · similarity 0.90)
- Geometric-Aware Hypergraph Reasoning for Novel Class Discovery in Point Cloud Segmentation (CVPR 2026 · similarity 0.89)
- Novel Class Discovery for Point Cloud Segmentation via Joint Learning of Causal Representation and Reasoning (NeurIPS 2025 · similarity 0.89)
- Weakly Supervised Point Cloud Semantic Segmentation via Artificial Oracle (CVPR 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.
