HSS-IAD
收藏资源简介:
HSS-IAD数据集是由复旦大学工程学院和计算机科学技术学院共同构建的,包含8590张金属类工业零件图像和精确的异常标注。这些零件在结构或外观上存在差异,但属于同一类产品,且缺陷细微,与基材相似,易于与加工痕迹、油污等混淆。数据集旨在为多类工业异常检测算法提供更具挑战性的基准,以促进算法在现实工厂条件下的性能提升。
The HSS-IAD dataset was jointly constructed by the School of Engineering and the School of Computer Science and Technology of Fudan University. It comprises 8,590 images of metal industrial components along with precise anomaly annotations. These parts belong to the same product category but differ in structure and appearance, with subtle defects that are highly similar to the base material and easily misidentified as processing marks, oil stains and other similar surface artifacts. This dataset aims to provide a more challenging benchmark for multi-class industrial anomaly detection algorithms, so as to promote the performance improvement of such algorithms under real-world factory conditions.




