Defect Spectrum
收藏资源简介:
Defect Spectrum数据集是由香港科技大学(广州)和SmartMore联合创建的,旨在提供精确、语义丰富的大规模工业缺陷标注。该数据集包含3518张高质量图像,来源于MVTEC、VISIONV1等四个关键工业基准,通过严格的标注过程,提供了详细的缺陷类型和位置信息。数据集不仅用于提升缺陷检测模型的性能,还支持视觉语言模型的研究。创建过程中,使用了创新的标注工具Defect-Click,以提高标注效率和准确性。该数据集主要应用于工业制造中的缺陷检测,旨在解决现有数据集在缺陷类型和位置标注上的不足。
Defect Spectrum Dataset was jointly created by The Hong Kong University of Science and Technology (Guangzhou) and SmartMore, aiming to provide precise and semantically rich large-scale industrial defect annotations. This dataset contains 3518 high-quality images sourced from four key industrial benchmarks including MVTEC and VISIONV1, and provides detailed defect type and location information through a rigorous annotation process. It can be used not only to enhance the performance of defect detection models, but also to support research on vision-language models. During the creation process, an innovative annotation tool Defect-Click was employed to improve annotation efficiency and accuracy. Primarily applied to defect detection in industrial manufacturing, this dataset aims to address the shortcomings of existing datasets in terms of defect type and location annotation.




