MAGIC: Few-Shot Mask-Guided Anomaly Inpainting with Prompt Perturbation, Spatially Adaptive Guidance, and Context Awareness
MAGIC: 少样本掩码引导的异常修复与提示扰动、空间自适应引导和上下文意识
机构 * Department of AI Semiconductor Engineering, Hanyang University, Seoul, Korea(汉阳大学人工智能半导体工程系) ; Department of Electronic Engineering, Hanyang University, Seoul, Korea(汉阳大学电子工程系)
专题命中 图像修复 :inpainting(title,abstract);diffusion(abstract);分类 cs.CV
AI总结 MAGIC通过引入提示扰动、空间自适应引导和上下文意识,解决少样本异常生成中的样本多样性问题,提升下游任务的鲁棒性。
Comments Accepted at CVPR 2026 Findings. Supplementary material included after references. 47 pages, 47 figures, 28 tables. Code : https://github.com/SpatialAILab/MAGIC