Research map / Label-efficient and robust learning / OOD and anomaly detection
Self-supervised Feature Adaptation for 3D Industrial Anomaly Detection
Yuanpeng Tu, Boshen Zhang, Liang Liu, YUXI LI, Jiangning Zhang, Yabiao Wang, Chengjie Wang, cairong zhao
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 Intra-modal Feature Compactness (IFC) optimization (ablation) | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC: 0.929 -> 0.959 | Improved |
| add Cross-modal Local-to-global Consistency (CLC) alignment (ablation) | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC: 0.929 -> 0.957 | Improved |
| combine IFC and CLC (ablation) | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC: 0.959 -> 0.971 | Improved |
| add global alignment loss L_GA alongside local alignment L_LA (ablation) | CLC · MVTec-3D AD (RGB+3D) | I-AUROC: 0.949 -> 0.959 | Improved |
| add global compactness loss L_GC alongside local L_LC (ablation) | IFC · MVTec-3D AD (RGB+3D) | I-AUROC: 0.952 -> 0.957 | Improved |
| use a single transformer encoder layer adaptor instead of two layers (ablation) | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC: 0.954 -> 0.974 | Improved |
| train adaptors instead of fine-tuning the whole backbone (ablation) | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC: 84.57 -> 95.42 | Improved |
| replace adaptors with LoRA or AdaLoRA (ablation) | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC: 97.1 -> 93.91 | Hurt |
| reduce training data to 5-shot (ablation) | LSFA · MVTec-3D AD few-shot | I-AUROC: 0.971 -> 0.834 | Hurt |
| vary memory bank size and balancing lambda (ablation) | LSFA · MVTec-3D AD | I-AUROC | No clear effect |
| LSFA vs AST on RGB-only | LSFA · MVTec-3D AD (RGB only) | I-AUROC mean: 0.880 -> 0.861 | Hurt |
| LSFA over M3DM on RGB+3D | LSFA · MVTec-3D AD (RGB+3D) | I-AUROC mean: 0.945 -> 0.971 | Improved |
Closest papers by meaning
- Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping (CVPR 2024 · similarity 0.90)
- FIND: Few-Shot Anomaly Inspection with Normal-Only Multi-Modal Data (ICCV 2025 · similarity 0.90)
- Real-IAD D3: A Real-World 2D/Pseudo-3D/3D Dataset for Industrial Anomaly Detection (CVPR 2025 · similarity 0.89)
- Rethinking Reverse Distillation for Multi-Modal Anomaly Detection (AAAI 2024 · similarity 0.88)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.87)
- 3CAD: A Large-Scale Real-World 3C Product Dataset for Unsupervised Anomaly Detection (AAAI 2025 · 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.
