DecurtainNet Dataset

Le résumé fourni par la source

DecurtainNet Dataset is an extensive collection of over 22,000 electron microscopy (EM) images meticulously curated to develop and validate DecurtainNet, a UNet-based deep learning model aimed at automatically removing curtaining artefact revealed in EM images. Curtaining artefacts, resulting from uneven milling rates during Focused Ion Beam (FIB) processing—particularly in composite or beam-sensitive materials—can significantly hinder accurate EM data interpretation. This dataset encompasses images from a diverse range of EM techniques and sample types, ensuring broad applicability and robustness of the model. Each image has undergone manual processing to correct for curtaining artefact patterns, providing high-quality training data for supervised learning approaches. By making this dataset publicly available on Figshare, we aim to support the advancement of artefact correction tools within the EM community and facilitate further research into enhancing the clarity and accuracy of EM analyses.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.