合成缺陷图像数据集

该数据通过 AIGC 技术生成工业产品的罕见缺陷样本图像，涵盖电子元件、机械部件、建筑材料等多类产品的低概率缺陷（如电子芯片引脚断裂、轴承滚子偏心、瓷砖边角崩裂等），每幅图像标注缺陷类型、位置、严重程度及生成参数，与真实缺陷数据结合后可扩大缺陷覆盖范围，提升检测模型对罕见缺陷的识别能力。This dataset generates rare defect sample images of industrial products via AIGC technology. It covers low-probability defects across multiple product categories including electronic components, mechanical parts, building materials, etc., with examples like broken pins of electronic chips, eccentricity of bearing rollers, and chipped edges and corners of ceramic tiles. Each image is annotated with defect type, location, severity level and generation parameters. When combined with real-world defect datasets, it can expand the coverage of defect types and enhance the recognition performance of defect detection models for rare defects.

搜集汇总

背景与挑战

背景概述

该数据集由江苏粉社数字科技有限公司创建，通过AIGC技术生成工业产品的罕见缺陷图像样本，涵盖电子芯片引脚断裂等低概率缺陷，并标注类型和位置信息；旨在与真实数据结合，提升检测模型对罕见缺陷的识别能力，数据以excel格式存储。