Research map / Detection and segmentation / Open-set and few-shot detection
Attention-Conditioned Augmentations for Self-Supervised Anomaly Detection and Localization
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 |
|---|---|---|---|
| use attention-conditioned patch masking rather than no masking (ablation) | ViT-S/16 pre-training on MVTec AD · MVTec AD | pixel-wise AUROC: 96.2 -> 98.2 | Helped |
| use attention-conditioned patch masking rather than random masking (ablation) | ViT-S/16 pre-training on MVTec AD · MVTec AD | pixel-wise AUROC: 97.1 -> 98.2 | Helped |
| replace DINO pre-training with APMask pre-training (ablation) | ViT-S/16 · MVTec AD | pixel-wise AUROC: 96.9±0.9 -> 98.2±0.4 | Helped |
| replace iBOT pre-training with APMask pre-training (ablation) | ViT-S/16 · MVTec AD | image-level AUROC: 97.1±0.9 -> 98.3±0.2 | Helped |
| replace MSN pre-training with APMask pre-training (ablation) | ViT-S/16 · MVTec AD | pixel-wise AUROC: 97.8±0.6 -> 98.2±0.4 | Helped |
| use attention-guided synthetic anomalies instead of CutPaste (ablation) | fine-tuning on MVTec AD · MVTec AD | pixel-wise AUROC: 96.1±1.1 -> 98.2±0.4 | Helped |
| use attention-guided synthetic anomalies instead of NSA (ablation) | fine-tuning on MVTec AD · MVTec AD | image-level AUROC: 97.3±0.5 -> 98.3±0.2 | Helped |
| use attention-guided synthetic anomalies instead of FPI (ablation) | fine-tuning on MVTec AD · MVTec AD | pixel-wise AUROC: 92.2±0.7 -> 98.2±0.4 | Helped |
| set APMask ratio to r = 0.15 rather than r = 0.3 (ablation) | ViT-S/16 pre-training · MVTec AD | pixel-wise AUROC: 97.9 -> 98.2 | Helped |
| set APMask ratio to r = 0.5 rather than r = 0.3 (ablation) | ViT-S/16 pre-training · MVTec AD | pixel-wise AUROC: 98.2 -> 97.8 | Hurt |
| use the proposed method rather than PII | NIH chest X-rays, male patients · NIH chest X-rays, male | pixel-level AUROC: 65.2±0.9 -> 70.9±0.5 | Helped |
| use the proposed method rather than PII | NIH chest X-rays, female patients · NIH chest X-rays, female | pixel-level AUROC: 65.1±1.1 -> 70.5±0.8 | Helped |
Closest papers by meaning
- Self-Supervised Predictive Convolutional Attentive Block for Anomaly Detection (CVPR 2022 · similarity 0.88)
- Self-Supervised Likelihood Estimation with Energy Guidance for Anomaly Segmentation in Urban Scenes (AAAI 2024 · similarity 0.85)
- Self-Supervised Learning of Intertwined Content and Positional Features for Object Detection (ICML 2025 · similarity 0.85)
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.84)
- Promptable Anomaly Segmentation with SAM Through Self-Perception Tuning (AAAI 2025 · similarity 0.84)
- Self-Guided Masked Autoencoder (NeurIPS 2024 · similarity 0.84)
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.