Research map / Label-efficient and robust learning / OOD and anomaly detection
GroundingAnomaly: Spatially-Grounded Diffusion for Few-Shot Anomaly Synthesis
Yishen Liu, Hongchang Chen, Pengcheng Zhao, Yunfan Bao, Yuxi Tian, Jieming Zhang, HAO CHEN, Zhi Zheng, Yongchun Liu, Ying Li, Dongpu Cao
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 Gated Self-Attention Module (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 75.7 -> 85.9 | Improved |
| add Spatial-Textual Feature Fusion (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 77.3 -> 85.9 | Improved |
| add Disentangled Token Learning (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 81.7 -> 85.9 | Improved |
| add mixed normal-anomalous training (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 83.1 -> 85.9 | Improved |
| add normal-prior denoising initialization (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 84.7 -> 85.9 | Improved |
| add mixed normal-anomalous training (ablation) | GroundingAnomaly · MVTec AD generation | IS: 1.74 -> 1.99 | Improved |
| add Gated Self-Attention Module (ablation) | GroundingAnomaly · VisA (U-Net) | pixel AP: 58.1 -> 67.2 | Improved |
| use all anomaly images (1/3) instead of 1-shot (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 72.3 -> 85.9 | Improved |
| use all anomaly images (1/3) instead of 1-shot (ablation) | GroundingAnomaly · VisA (U-Net) | pixel AP: 44.3 -> 67.2 | Improved |
| use N=4, K=4 tokens instead of N=1, K=1 (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 80.5 -> 85.9 | Improved |
| use gated self-attention instead of ungated (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 74.1 -> 85.9 | Improved |
| add LoRA to self-attention layers (ablation) | GroundingAnomaly · MVTec AD (U-Net) | pixel AP: 83.3 -> 85.9 | Improved |
Closest papers by meaning
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.91)
- One-to-More: High-Fidelity Training-Free Anomaly Generation with Attention Control (CVPR 2026 · similarity 0.91)
- Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation (CVPR 2025 · similarity 0.90)
- SeaS: Few-shot Industrial Anomaly Image Generation with Separation and Sharing Fine-tuning (ICCV 2025 · similarity 0.90)
- FAST: Foreground‑aware Diffusion with Accelerated Sampling Trajectory for Segmentation‑oriented Anomaly Synthesis (NeurIPS 2025 · similarity 0.89)
- Unseen Visual Anomaly Generation (CVPR 2025 · similarity 0.88)
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.
