LU5M812TGT: An AI-Powered Global Database of impact craters >=0.4 km on the Moon.

This dataset presents the results of a groundbreaking AI-driven lunar crater mapping project, marking the first comprehensive application of artificial intelligence to detect, classify, and map lunar craters. Created through extensive work over a two-year period, this dataset leverages YOLOLens, a state-of-the-art deep learning model specifically optimized for high-resolution crater detection. YOLOLens, an innovative variant of the YOLO architecture, has been fine-tuned to handle the unique challenges of lunar surface imagery, delivering unparalleled accuracy in crater identification and localization. Detailed information on the model architecture and methodology can be found in relevant publications on: La Grassa, Riccardo, et al. "YOLOLens: A deep learning model based on super-resolution to enhance the crater detection of the planetary surfaces." Remote Sensing 15.5 (2023): 1171, https://doi.org/10.3390/rs15051171. La Grassa, R, et al. "LU5M812TGT: An AI-Powered global database of impact craters ≥0.4 km on the Moon", ISPRS Journal of Photogrammetry and Remote Sensing, 2025, ISSN 0924-2716, https://doi.org/10.1016/j.isprsjprs.2024.11.010. La Grassa, Riccardo, et al. "From the Moon to Mercury: Release of Global Crater Catalogs Using Multimodal Deep Learning for Crater Detection and Morphometric Analysis" Remote Sensing 17.19 (2025): 3287. NEW: We have released a plugin available on the official repositor of QGIS, please check also the documentation and pretrained models -> https://github.com/riccardolagrassa/YOLOLens_QGIS_Plugin The dataset preparation process involved rigorous steps in preprocessing and post-processing to enhance the quality and usability of the data. Preprocessing techniques were employed to reduce noise and enhance contrast within the complex lunar landscape, while post-processing was used to refine the accuracy of crater boundaries and dimensions detected by the model. This approach facilitated a high-confidence dataset that stands as a valuable resource for the astronomical and AI research communities. This dataset can serve as a critical tool for both astrophysicists and AI researchers. By providing labelled crater data with precise coordinates, dimensions, and classifications, it offers a benchmark for comparative studies, model validation, and further innovations in celestial object detection. *********************************************************************************************************************************************************** Release of a new global craters catalog (LU5M812TGT) with more than 5 million craters. File source csv Header: Longitude, Latitude, Diameter_w, Diameter_h, Confidence. The coordinates are absolute in the range of [-180, +180] of Longitude and [-90, +90] of Latitude. Diameters (km)本数据集呈现了一项具有开创性的AI驱动月球陨石坑测绘项目成果，这是人工智能首次被全面应用于月球陨石坑的检测、分类与测绘工作。本数据集历经两年的系统性研究工作构建完成，依托YOLOLens——一款专为高分辨率陨石坑检测优化的前沿深度学习模型。YOLOLens是YOLO架构的创新变体，经过微调以应对月球表面影像的独特挑战，在陨石坑识别与定位方面实现了前所未有的精度。有关模型架构与研究方法的详细信息可参阅以下发表文献： 1. La Grassa, Riccardo 等：《YOLOLens：基于超分辨率的深度学习模型，用于提升行星表面陨石坑检测能力》，《遥感》，2023年，第15卷第5期，第1171页，https://doi.org/10.3390/rs15051171。 2. La Grassa, R 等：《LU5M812TGT：月球上≥0.4公里撞击坑的AI驱动全球数据库》，《国际摄影测量与遥感学会摄影测量与遥感期刊》，2025年，ISSN 0924-2716，https://doi.org/10.1016/j.isprsjprs.2024.11.010。 3. La Grassa, Riccardo 等：《从月球到水星：利用多模态深度学习实现陨石坑检测与形态计量分析的全球陨石坑目录发布》，《遥感》，2025年，第17卷第19期，第3287页。 新动态：我们已在QGIS官方仓库发布一款插件，敬请查阅相关文档与预训练模型：https://github.com/riccardolagrassa/YOLOLens_QGIS_Plugin 本数据集的制备过程包含严谨的预处理与后处理步骤，以提升数据质量与可用性。预处理技术用于降低复杂月球地貌中的噪声并增强对比度，后处理则用于优化模型检测到的陨石坑边界与尺寸精度。该方法打造出了高置信度的数据集，可为天文学与AI研究领域提供宝贵资源。 本数据集可作为天体物理学家与AI研究人员的关键工具。通过提供包含精确坐标、尺寸与分类的标注陨石坑数据，它为天体目标检测领域的对比研究、模型验证及后续创新提供了可靠基准。 *********************************************************************************************************************************************************** 全新发布：包含超过500万个陨石坑的全球陨石坑目录（LU5M812TGT）。 文件格式：CSV 表头字段：经度（Longitude）、纬度（Latitude）、长轴直径（Diameter_w）、短轴直径（Diameter_h）、置信度（Confidence）。 坐标为绝对坐标，经度范围为[-180, +180]，纬度范围为[-90, +90]。 直径单位：千米（km）