Research map / Continual learning / Class-incremental
An Incremental Unified Framework for Small Defect Inspection
Jiaqi Tang, Hao Lu, Xiaogang Xu, Ruizheng Wu, Sixing Hu, Tong Zhang, Tsz Wa Cheng, Ming Ge, Ying-Cong Chen, Fugee Tsung
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 Object-Aware Self-Attention to the reconstruction network (ablation) | IUF · MVTec-AD 10-1 with 5 Steps | image-level ACC: 84.2 -> 94.2 | Improved |
| add updating strategy that retains old weights and suppresses updates in primary old semantic space (ablation) | IUF · MVTec-AD 10-1 with 5 Steps | image-level ACC: 85.6 -> 94.2 | Improved |
| add Semantic Compression Loss (ablation) | IUF · MVTec-AD 10-1 with 5 Steps | image-level ACC: 93.4 -> 94.2 | Improved |
| add Semantic Compression Loss (ablation) | IUF · MVTec-AD 10-5 with 1 Step | image-level FM: 8.1 -> 9.3 | Hurt |
| add Object-Aware Self-Attention (ablation) | IUF · MVTec-AD 14-1 with 1 Step | image-level ACC: 89.3 -> 96.0 | Improved |
| add updating strategy (ablation) | IUF · MVTec-AD 14-1 with 1 Step | image-level ACC: 95.9 -> 96.0 | No clear effect |
| use IUF instead of UniAD baseline | UniAD · MVTec-AD 10-1 with 5 Steps | image-level FM: 21.1 -> 3.2 | Improved |
| use IUF instead of UniAD + EWC in 3-3 with 4 steps | UniAD · MVTec-AD 3-3 with 4 Steps | image-level FM: ACC 81.3 -> 84.2 but FM 7.4 -> 10.0 | Mixed |
| use IUF instead of CAD-based methods (memory bank) | incremental defect inspection · MVTec-AD 14-1 with 1 Step | image-level ACC: 84.5 -> 96.0 | Improved |
| use IUF instead of UniAD + SI on VisA 8-4 with 1 step | UniAD · VisA 8-4 with 1 Step | image-level ACC: 80.8 -> 80.1 | Hurt |
Closest papers by meaning
- Towards Open-Vocabulary Industrial Defect Understanding with a Large-Scale Multimodal Dataset (CVPR 2026 · similarity 0.86)
- InCLAD: A Continual Learning Benchmark for Industrial Visual Anomaly Detection (NeurIPS 2026 · similarity 0.86)
- MVREC: A General Few-shot Defect Classification Model Using Multi-View Region-Context (AAAI 2025 · similarity 0.85)
- Wavelet and Prototype Augmented Query-based Transformer for Pixel-level Surface Defect Detection (CVPR 2025 · similarity 0.84)
- Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection (AAAI 2024 · similarity 0.83)
- Zero-Shot Learning in Industrial Scenarios: New Large-Scale Benchmark, Challenges and Baseline (AAAI 2025 · similarity 0.83)
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.
