<b>CoatingVision: A Defect Dataset for Coating Manufacturing</b>

Electrode is a key component of many energy storage and energy conversion devices such as batteries and fuel cells. Defects in electrodes can significantly influence device performance and reliability and thus need to be monitored and eliminated during the electrode manufacturing process. Advancements in in-line metrology, computer vision, and machine learning have enabled the development of integrated hardware-software systems for automated defect detection and diagnostics. While several manufacturing domains have published defect datasets to support such efforts, publicly available datasets specific to electrode coating processes are not available. To fill this gap and support research on defect detection for automated coating processes, we present CoatingVision, a comprehensive dataset of slot-die coating images with labeled defect types. This dataset supports a diverse range of image recognition tasks, including defect segmentation, defect detection, and multi-label classification. It includes high-resolution images with associated labels for common defects such as surface cracks, delamination cracks, pinholes, and unclassified defects. To facilitate benchmarking and reproducible research, CoatingVision is packaged with an open-source codebase that enables comparative evaluation of AI models and hyperparameter configurations. The dataset has been meticulously curated to ensure high quality and consistency, providing researchers with reliable data for training and evaluating computer vision models. With over 2,200 image samples under various processing conditions, CoatingVision offers a robust foundation for developing automated defect detection systems. It promotes deeper insights into defect formation in coating manufacturing processes, which can be used to advance various coating-related applications including batteries and fuel cells.

电极（Electrode）是各类储能与能量转换器件（如电池、燃料电池）的核心组件。电极缺陷会显著影响器件的性能与可靠性，因此在电极制造流程中需对其进行监测并加以消除。在线计量（in-line metrology）、计算机视觉与机器学习领域的技术进步，推动了集成化软硬件系统的开发，可实现自动化缺陷检测与诊断。尽管多个制造领域已公开相关缺陷数据集以支撑此类研究，但针对电极涂布工艺的公开可用数据集仍较为稀缺。为填补这一研究空白，并支撑自动化涂布工艺的缺陷检测相关研究，我们推出了CoatingVision数据集——一款带有标注缺陷类型的狭缝涂布（slot-die coating）图像综合数据集。该数据集可支撑多样化的图像识别任务，涵盖缺陷分割、缺陷检测与多标签分类。其包含高分辨率图像，并附带常见缺陷的标注标签，包括表面裂纹、分层裂纹、针孔以及未分类缺陷。为便于基准测试与可复现研究，CoatingVision配套了开源代码库，可实现AI模型与超参数配置的对比评估。本数据集经过精心整理，以确保高质量与一致性，为研究人员训练与评估计算机视觉模型提供了可靠的数据支撑。该数据集包含超过2200张不同工艺条件下的图像样本，为开发自动化缺陷检测系统提供了坚实的基础。它有助于深入理解涂布制造过程中的缺陷形成机制，可用于推动包括电池与燃料电池在内的各类涂布相关应用的发展。