研究人员开发了 SynSur，一个用于生成合成工业表面缺陷的端到端管线，以解决缺陷检测中标记数据稀缺的问题。该管线结合了视觉语言模型、LoRA 适配的扩散模型和掩码引导的图像修复技术，以创建逼真的缺陷样本。实验表明，虽然合成数据本身不能替代真实数据，但当与现有数据集结合使用时，可以提高性能，尤其是在改进训练方案和跨域迁移方面。 AI
影响 通过用逼真的合成样本扩充稀缺的真实世界数据集，增强了工业缺陷检测能力。
排序理由 该集群描述了一篇学术论文，详细介绍了一种新的合成数据生成方法。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
The bottleneck in learning-based industrial defect detection is often limited not by model capacity, but by the scarcity of labeled defect data: defects are rare, annotations are expensive, and collecting balanced training sets is slow. We present an end-to-end pipeline for synth…
arXiv:2604.26633v1 Announce Type: new Abstract: The bottleneck in learning-based industrial defect detection is often limited not by model capacity, but by the scarcity of labeled defect data: defects are rare, annotations are expensive, and collecting balanced training sets is s…
The bottleneck in learning-based industrial defect detection is often limited not by model capacity, but by the scarcity of labeled defect data: defects are rare, annotations are expensive, and collecting balanced training sets is slow. We present an end-to-end pipeline for synth…