Research map / Label-efficient and robust learning / OOD and anomaly detection
CASL: Curvature-Augmented Self-supervised Learning for 3D Anomaly Detection
Yaohua Zha, Xue Yuerong, Chunlin Fan, Yuansong Wang, Tao Dai, Ke Chen, Shu-Tao Xia
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 |
|---|---|---|---|
| CASL over best prior (ISMP) | ISMP · Real3D-AD | O-AUROC: 0.767 -> 0.823 | Improved |
| CASL over PO3AD | PO3AD · Anomaly-ShapeNet | O-AUROC: 0.839 -> 0.887 | Improved |
| CASL over PO3AD | PO3AD · Anomaly-ShapeNet | P-AUROC: 0.898 -> 0.899 | No clear effect |
| curvature alone as anomaly score vs PatchCore | PatchCore · Real3D-AD | O-AUROC: 0.682 -> 0.723 | Improved |
| curvature alone as anomaly score vs PO3AD | PO3AD · Anomaly-ShapeNet | O-AUROC: 0.839 -> 0.559 | Hurt |
| CASL vs PointGPT-S | PointGPT-S · ScanObjectNN OBJ-BG | accuracy (%): 91.60 -> 92.08 | Improved |
| CASL vs PointGPT-S | PointGPT-S · ScanObjectNN PB-T50-RS | accuracy (%): 86.90 -> 86.81 | Hurt |
| CASL vs PointGPT-S | PointGPT-S · ShapeNetPart | mIoU_I (%): 86.2 -> 86.2 | No clear effect |
| add 60% masking to coordinate self-reconstruction (ablation) | coordinate-only encoder · Real3D-AD | O-AUROC: 0.751 -> 0.776 | Improved |
| replace coordinates with curvature input (no masking) (ablation) | coordinate-only encoder · Real3D-AD | P-AUROC: 0.649 -> 0.878 | Improved |
| add curvature prompts and full masking to curvature-only (ablation) | Experiment C · Real3D-AD | O-AUROC: 0.804 -> 0.823 | Improved |
| raise masking ratio to 100% (ablation) | CASL · Real3D-AD | AUROC | Improved |
Closest papers by meaning
- Towards Scalable 3D Anomaly Detection and Localization: A Benchmark via 3D Anomaly Synthesis and A Self-Supervised Learning Network (CVPR 2024 · similarity 0.90)
- Registration is a Powerful Rotation-Invariance Learner for 3D Anomaly Detection (NeurIPS 2025 · similarity 0.87)
- Towards Generalizable 3D Anomaly Detection via Relational Inconsistency Modeling (NeurIPS 2026 · similarity 0.86)
- PO3AD: Predicting Point Offsets toward Better 3D Point Cloud Anomaly Detection (CVPR 2025 · similarity 0.85)
- Looking 3D: Anomaly Detection with 2D-3D Alignment (CVPR 2024 · similarity 0.85)
- Real-IAD D3: A Real-World 2D/Pseudo-3D/3D Dataset for Industrial Anomaly Detection (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.
