Research map / Image generation / Autoregressive and latent generation
UniScale: Arbitrary-Scale Anomaly Generation
Shilei Zeng, Linxin Guan, Xurui Li, Yaohan Tang, Yu Zhou
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 |
|---|---|---|---|
| replace SeaS with UniScale generated data | segmentation model · VisA | IS(a): 1.81 -> 2.64 | Improved |
| replace SeaS with UniScale generated data | segmentation model · MVTec AD 2 | pixel-level AUROC: 71.12 -> 77.67 | Improved |
| replace SeaS with UniScale generated data | segmentation model · VisA | pixel-level IoU: 25.93 -> 30.15 | Improved |
| add Error-Suppressed Multi-Scale Training (multi-scale crops + CSP) (ablation) | UniScale · VisA | IS(a): 1.84 -> 2.64 | Improved |
| add Cross-Scale Perception loss (ablation) | UniScale · VisA | IS(a): 2.01 -> 2.64 | Improved |
| add Generation-then-Fusion Denoising (ablation) | UniScale · VisA | pixel-level AP: 33.62 -> 42.91 | Improved |
| add maximum-scale crop (ablation) | UniScale · VisA | pixel-level AP: 33.62 -> 42.91 | Improved |
| add minimum-scale crop (ablation) | UniScale · VisA | IS(a): 2.04 -> 2.64 | Improved |
| add intermediate-scale crops (ablation) | UniScale · VisA | pixel-level AUROC: 93.48 -> 97.28 | Improved |
| crop expansion factor alpha 4 vs 16 (ablation) | UniScale · VisA | IS(a): 2.34 -> 2.64 | Improved |
| crop expansion factor alpha 4 vs 2 (ablation) | UniScale · VisA | IS(a): 2.63 -> 2.64 | No clear effect |
| progressive multi-scale fusion vs direct paste (ablation) | UniScale · VisA | pixel-level IoU: 27.65 -> 30.15 | Improved |
Closest papers by meaning
- Few-Shot Anomaly-Driven Generation for Anomaly Classification and Segmentation (ECCV 2024 · similarity 0.91)
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.90)
- Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation (CVPR 2025 · similarity 0.90)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.88)
- Unseen Visual Anomaly Generation (CVPR 2025 · similarity 0.88)
- FAST: Foreground‑aware Diffusion with Accelerated Sampling Trajectory for Segmentation‑oriented Anomaly Synthesis (NeurIPS 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.
