PEMFuelCell-Defect1934

<b>PEMFuelCell-Defect1934</b> is a curated grayscale image dataset for <b>single-class defect localization</b> in proton-exchange membrane (PEM) fuel cell materials. The dataset is intended for <b>bounding-box–based defect detection</b> and supports research in <b>materials AI, data-centric machine learning, and computer vision</b>.The dataset contains <b>1,934 images</b> in total and is organized under a single root directory (<code>PEMFuelCell-Defect1934</code>). Images are split into <b>1,872 training images</b>, <b>31 validation images</b>, and <b>31 test images</b>, with each split stored in a dedicated subfolder under <code>images/</code>. All corresponding bounding-box annotations are provided as JSON files in the <code>all_annotations/</code> folder, with separate annotation files for the training, validation, and test sets.All images are provided in <b>PNG format</b>, resized to a fixed resolution of <b>640 × 640</b>, and stored in <b>grayscale</b>. A physics-consistent preprocessing pipeline was applied uniformly across the dataset, including conversion from proprietary microscopy formats, deterministic removal of scale bars via cropping, and global intensity normalization. Scale bars were removed to prevent shortcut learning from acquisition metadata and to ensure that models learn from physical structure rather than imaging artifacts.The train/validation/test split was performed <b>before any data augmentation</b>, ensuring strict separation between splits. Validation and test images are <b>not augmented</b>, enabling unbiased and reproducible evaluation of defect localization performance.This dataset is suitable for training and benchmarking defect localization models in PEM fuel cell imagery and for studying data-centric effects in materials-focused computer vision.