Research map / Detection and segmentation / Open-set and few-shot detection
DeSTSeg: Segmentation Guided Denoising Student-Teacher for Anomaly Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train the vanilla student on synthetic anomalous inputs without changing its architecture (ablation) | Student-teacher anomaly detection · MVTec AD | image-level AUC (%): 94.8 -> 93.4 | Hurt |
| Pair denoising training with an encoder-decoder student (ablation) | Student-teacher anomaly detection · MVTec AD | pixel-level AP (%): 52.9 -> 54.0 | Helped |
| Add a segmentation network for learned multi-level feature fusion (ablation) | Student-teacher anomaly detection · MVTec AD | pixel-level AP (%): 52.9 -> 70.1 | Helped |
| Add denoising training to the encoder-decoder student and segmentation network (ablation) | DeSTSeg · MVTec AD | pixel-level AP (%): 69.7 -> 75.8 | Helped |
| Add L1 loss to the segmentation objective (ablation) | DeSTSeg · MVTec AD | pixel-level AP (%): 72.2 -> 75.8 | Helped |
| Use element-wise products of normalized student-teacher features instead of concatenated features (ablation) | DeSTSeg segmentation network · MVTec AD | pixel-level AP (%): 72.2 -> 75.8 | Helped |
| Use element-wise products of normalized student-teacher features instead of cosine-distance features (ablation) | DeSTSeg segmentation network · MVTec AD | pixel-level AP (%): 72.0 -> 75.8 | Helped |
| Use DeSTSeg rather than the strongest reported baseline | MVTec AD image-level anomaly detection · MVTec AD | image-level AUC (%): 98.5 -> 98.6 | Helped |
| Use DeSTSeg rather than PatchCore | MVTec AD pixel-level localization · MVTec AD | pixel-level AUC (%): 98.4 -> 97.9 | Mixed |
| Use DeSTSeg rather than the strongest reported baseline | MVTec AD pixel-level localization · MVTec AD | pixel-level AP (%): 70.2 -> 75.8 | Helped |
| Use DeSTSeg rather than DRAEM | MVTec AD instance-level detection · MVTec AD | instance-level IAP (%): 71.5 -> 76.4 | Helped |
| Use DeSTSeg rather than DRAEM | MVTec AD high-recall instance detection · MVTec AD | instance-level IAP@90 (%): 51.7 -> 57.8 | Helped |
Closest papers by meaning
- A Diffusion-Based Framework for Multi-Class Anomaly Detection (AAAI 2024 · similarity 0.87)
- CostFilter-AD: Enhancing Anomaly Detection through Matching Cost Filtering (ICML 2025 · similarity 0.86)
- Unlocking the Potential of Reverse Distillation for Anomaly Detection (AAAI 2025 · similarity 0.85)
- Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · similarity 0.85)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.85)
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 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.