Taktpixel2025PD-CD: Taktpixel 2025 Printing Defect Change Detection Dataset

Taktpixel2025PD-CD is a dataset for evaluating change detection models in the domain of industrial printing inspection. It contains image pairs of printed materials scanned from both offset and inkjet print samples. Each image pair includes: A master image (reference, assumed defect-free) A target image (possibly containing a defect) A pixel-wise binary mask representing the defect region The dataset is divided into train, val, and test splits, and organized into the standard A/, B/, and OUT/ folder structure commonly used in change detection tasks. Six common defect types are labeled: Black spot Color shift Friction Hair Line Pinhole Candidate defect regions were extracted using comparative image inspection techniques and manually validated through expert review. Ground-truth masks are refined to exclude false positives caused by normal printing variation. This dataset is released under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license. It is suitable for training and benchmarking machine learning models, particularly for applications in visual inspection and quality assurance in the printing industry.Taktpixel2025PD-CD 是一款面向工业印刷检测领域变化检测模型的评估数据集。该数据集收录了源自胶印与喷墨印刷样本的印刷品扫描图像对，每组图像对均包含三类内容：一是主图像（参考图像，预设无缺陷），二是目标图像（可能存在缺陷），三是用于标记缺陷区域的逐像素二值掩码。该数据集划分为训练集、验证集与测试集，并采用变化检测任务中通用的标准A/、B/及OUT/文件夹结构进行组织。数据集共标注了6类常见印刷缺陷，分别为黑点、色偏、摩擦、发丝、线条与针孔。候选缺陷区域通过对比图像检测技术提取，并经由专家评审完成人工验证；真值掩码已经过精修，以排除由正常印刷波动引发的假阳性样本。本数据集采用知识共享署名-相同方式共享4.0国际（Creative Commons Attribution-ShareAlike 4.0 International，CC BY-SA 4.0）协议发布。该数据集可用于机器学习模型的训练与基准测试，尤其适用于印刷行业的视觉检测与质量保障相关应用。