Eyecandies
收藏资源简介:
Eyecandies数据集是由Eyecan.ai创建的一个新颖的合成数据集,专为无监督异常检测和定位设计。该数据集包含10类糖果的图像,每类糖果在工业传送带场景下,在多种光照条件下渲染而成,同时提供深度和法线图。数据集提供无缺陷样本用于模型训练和验证,而异常实例仅在测试集中提供,并附有精确的地面实况标注。该数据集旨在通过结合颜色、深度和法线图等额外信息,鼓励探索解决异常检测任务的新方法,特别是在工业质量保证等领域。
The Eyecandies Dataset is a novel synthetic dataset developed by Eyecan.ai, exclusively designed for unsupervised anomaly detection and localization tasks. It comprises images of 10 categories of candies, rendered under diverse lighting conditions within an industrial conveyor belt setting, paired with corresponding depth maps and normal maps. The dataset offers defect-free samples for model training and validation, whereas anomalous instances are exclusively present in the test split, together with accurate ground-truth annotations. This dataset is intended to facilitate the exploration of innovative solutions for anomaly detection tasks, particularly in domains such as industrial quality assurance, by leveraging supplementary information including color, depth and normal maps.




