Research map / Detection and segmentation / Open-set and few-shot detection
A Semantically Disentangled Unified Model for Multi-category 3D Anomaly Detection
SuYeon Kim, Wongyu Lee, MyeongAh Cho
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 |
|---|---|---|---|
| add CFGT to the baseline (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 75.1 -> 76.4 | Improved |
| add C3L on top of CFGT (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 76.4 -> 79.2 | Improved |
| add GGD on top of CFGT and C3L (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 79.2 -> 81.0 | Improved |
| add GGD on top of CFGT and C3L (ablation) | SeDiR · Real3D-AD (12 categories) ablation | P-AUROC: 79.0 -> 80.6 | Improved |
| add auxiliary classification loss to L_scl in C3L (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 79.1 -> 80.0 | Improved |
| add cosine alignment loss to L_scl and L_cls (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 80.0 -> 81.0 | Improved |
| use attention bias instead of masking in GGD (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 79.8 -> 81.0 | Improved |
| use attention bias instead of gating in GGD (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 79.7 -> 81.0 | Improved |
| use CFGT instead of ACT-only global token (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 78.5 -> 81.0 | Improved |
| use CFGT instead of max pooling global token (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 76.4 -> 81.0 | Improved |
| use CFGT instead of mean pooling global token (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 78.2 -> 81.0 | Improved |
| use ratios {k,2k,k/2} instead of {k,1.5k,2k/3} (ablation) | SeDiR · Real3D-AD (12 categories) ablation | O-AUROC: 79.9 -> 81.0 | Improved |
Closest papers by meaning
- Towards Generalizable 3D Anomaly Detection via Relational Inconsistency Modeling (NeurIPS 2026 · similarity 0.90)
- Towards Scalable 3D Anomaly Detection and Localization: A Benchmark via 3D Anomaly Synthesis and A Self-Supervised Learning Network (CVPR 2024 · similarity 0.89)
- PointAD: Comprehending 3D Anomalies from Points and Pixels for Zero-shot 3D Anomaly Detection (NeurIPS 2024 · similarity 0.87)
- Looking 3D: Anomaly Detection with 2D-3D Alignment (CVPR 2024 · similarity 0.87)
- Align3D-AD: Cross-Modal Feature Alignment and Dual-Prompt Learning for Zero-shot 3D Anomaly Detection (NeurIPS 2026 · similarity 0.86)
- PO3AD: Predicting Point Offsets toward Better 3D Point Cloud 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.
