Research map / Architectures and efficiency / RNNs and normalization
Revisiting Reverse Distillation 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 |
|---|---|---|---|
| Add RD++ projection layers and multi-task training objectives to RD | RD · MVTec | AUROC image-level (Avg.): 98.46 -> 99.44 | Helped |
| Add RD++ projection layers and multi-task training objectives to RD | RD · BTAD | AUROC image-level (Avg.): 94.30 -> 95.63 | Helped |
| Add RD++ projection layers and multi-task training objectives to RD | RD · BTAD | AUROC pixel-level (Avg.): 97.67 -> 97.43 | Hurt |
| Add RD++ projection layers and multi-task training objectives to RD | RD · BTAD | PRO (Avg.): 77.10 -> 77.30 | Helped |
| Add RD++ projection layers and multi-task training objectives to RD | RD · Retinal OCT | AUROC sample: 99.36 -> 99.73 | Helped |
| Add self-supervised optimal transport loss (ablation) | RD · MVTec | AUROC image-level: 98.46 -> 99.33 | Helped |
| Add contrast loss to the SSOT objective (ablation) | RD + SSOT · MVTec | PRO: 94.95 -> 95.00 | No clear effect |
| Add reconstruction loss to SSOT and contrast loss (ablation) | RD + SSOT + contrast loss · MVTec | PRO: 95.00 -> 94.99 | No clear effect |
| Set pseudo-anomaly noise level to 0.2 rather than 0.1 (ablation) | RD++ · MVTec | AUROC image-level: 98.98 -> 99.44 | Helped |
| Increase pseudo-anomaly noise level from 0.2 to 0.5 (ablation) | RD++ · MVTec | PRO: 94.99 -> 95.20 | Helped |
| Increase pseudo-anomaly noise level from 0.2 to 0.5 (ablation) | RD++ · MVTec | AUROC image-level: 99.44 -> 99.19 | Hurt |
| Use a wider and deeper WideResnet50 backbone rather than Resnet18 (ablation) | RD++ · MVTec | AUROC sample: 98.63 -> 99.44 | Helped |
Closest papers by meaning
- Unlocking the Potential of Reverse Distillation for Anomaly Detection (AAAI 2025 · similarity 0.88)
- Rethinking Reverse Distillation for Multi-Modal Anomaly Detection (AAAI 2024 · similarity 0.86)
- Co-Progression Knowledge Distillation with Knowledge Prototype for Industrial Anomaly Detection (AAAI 2025 · similarity 0.84)
- Scale-Aware Contrastive Reverse Distillation for Unsupervised Medical Anomaly Detection (ICLR 2025 · similarity 0.84)
- RealNet: A Feature Selection Network with Realistic Synthetic Anomaly for Anomaly Detection (CVPR 2024 · similarity 0.84)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · 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.