Research map / Label-efficient and robust learning / OOD and anomaly detection
SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation
Yang Zou, Jongheon Jeong, Latha Pemula, Dongqing Zhang, Onkar Dabeer
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 SPD regularization | SimSiam pre-training · VisA 1-class (PaDiM) classification | AU-PR: 80.2 -> 82.8 | Improved |
| add SPD regularization | SimSiam pre-training · VisA 2-class high-shot segmentation | AU-PR: 53.8 -> 59.7 | Improved |
| add SPD regularization | supervised pre-training · VisA 2-class high-shot segmentation | AU-PR: 65.1 -> 71.9 | Improved |
| add SPD regularization | MoCo pre-training · VisA 1-class (PaDiM) classification | AU-PR: 83.6 -> 84.1 | Improved |
| add SPD regularization | SimCLR pre-training · VisA 2-class high-shot classification | AU-PR: 93.4 -> 92.7 | Hurt |
| add SPD regularization | SimCLR pre-training · MVTec-AD 1-class (PaDiM) classification | AU-PR: 94.7 -> 96.8 | Improved |
| add SPD regularization | SimSiam pre-training · VisA 2-class 5-shot classification | AU-PR: 51.9 -> 56.1 | Improved |
| add SPD regularization | MoCo pre-training · VisA 2-class 5-shot segmentation | AU-ROC: 80.5 -> 78.5 | Hurt |
| use SPD loss weight 0.1 instead of 0.5 (ablation) | SimSiam+SPD · VisA 1-class (PaDiM) classification | AU-PR: 80.5 -> 82.8 | Improved |
| use SmoothBlend instead of CutPaste (ablation) | SimSiam+SPD · VisA 1-class (PaDiM) classification | AU-PR: 78.8 -> 82.8 | Improved |
| use SPD loss instead of cross-entropy loss (ablation) | SimSiam+SPD · VisA 1-class (PaDiM) classification | AU-PR: 71.4 -> 82.8 | Improved |
| add SPD regularization | supervised Wide ResNet50, PatchCore · VisA 1-class classification | AU-PR: 93.3 -> 93.8 | Improved |
Closest papers by meaning
- InCLAD: A Continual Learning Benchmark for Industrial Visual Anomaly Detection (NeurIPS 2026 · similarity 0.88)
- Supervised Anomaly Detection for Complex Industrial Images (CVPR 2024 · similarity 0.88)
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 2025 · similarity 0.87)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.85)
- MetaUAS: Universal Anomaly Segmentation with One-Prompt Meta-Learning (NeurIPS 2024 · similarity 0.85)
- RealNet: A Feature Selection Network with Realistic Synthetic Anomaly for Anomaly Detection (CVPR 2024 · 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.
