LU6M371TGT: A Global Lunar Crater Catalog derived by the multimodal YOLOLens network and enhanced with Morphometric Parameters

This repository hosts the <b>LU6M371TGT lunar crater catalog</b>, generated through a revised version of the YOLOLens model. By integrating visible imagery and elevation data (LOLA-derived DTMs), the model significantly enhances crater detection performance, particularly in regions with challenging illumination conditions such as the <b>Permanent Shadow Regions (PSRs)</b> at the lunar poles.Detailed information on the model architecture and methodology can be found in relevant publications on (see point 3 for the last <b>LU6M371TGT</b> catalog):<br><i>La Grassa, Riccardo, et al. </i><b><i>"YOLOLens: A deep learning model based on super-resolution to enhance the crater detection of the planetary surfaces."</i></b><i> Remote Sensing 15.5 (2023): 1171, https://doi.org/10.3390/rs15051171.</i><i>La Grassa, R, et al. </i><b><i>"LU5M812TGT: An AI-Powered global database of impact craters ≥0.4 km on the Moon"</i></b><i>, ISPRS Journal of Photogrammetry and Remote Sensing, 2025, ISSN 0924-2716, https://doi.org/10.1016/j.isprsjprs.2024.11.010.</i><i>La Grassa, R, et al. "</i><b><i>From the Moon to Mercury: Release of Global Crater Catalogs using Multimodal Deep Learning for Crater Detection and Morphometric Analysis</i></b><i>", </i><i>Remote Sens.</i> <b>2025</b>, <i>17</i>, 3287. https://doi.org/10.3390/rs17193287<br>The catalog contains <b>6,371,337 craters</b>, spanning diameters from <b>0.3 km to 109 km</b>, and provides the following variables for each entry:<b>Longitude, Latitude</b> – crater center coordinates in decimal degrees.<b>Diameter W, H</b> – horizontal and vertical bounding box dimensions (km).<b>Confidence</b> – model-assigned certainty score of crater identification.<b>Morphometric parameters</b> – depth estimates, rim elevations, and depth-to-diameter ratios derived from global DTMs.<b>Morphometric Extraction:</b><br>To enrich the AI-derived catalog with quantitative geomorphological measures, crater morphometry was extracted using a neighborhood-based sampling approach on high-resolution global DTMs. For each crater predicted, elevation profiles were computed by defining a spatial window proportional to the crater’s diameter, thereby minimizing bounding box misalignment. Key metrics include:<b>Central depth</b> – minimum elevation within the crater neighborhood.<b>Rim peak elevations </b>– maximum elevation values around crater rims.<b>Differential rim-to-center depths </b>– elevation difference between crater center and rim.<b>Depth-to-diameter ratio </b>– normalized depth as a function of average crater diameter.<b>Benchmarking and Validation:</b><br>The LU6M371TGT catalog was validated against established lunar crater datasets, including the Robbins Ground-Truth catalog and the previous <b>LU5M812TGT catalog</b>. The revised YOLOLens achieved <b>substantial recall improvements</b>, nearly doubling performance in polar regions. Manual validation by expert operators confirmed a <b>low false-positive rate</b> and robust generalization across diverse lunar terrains.<b>Key Features:</b>Global coverage of the Moon at high resolution.Improved crater detection in shaded and PSR-affected areas.Multimodal integration of WAC imagery and LOLA elevation data.Confidence-based filtering for customizable analysis.Enrichment with morphometric parameters for geomorphological research.<b>Applications:</b><br>This dataset supports studies in <b>lunar surface characterization, planetary geology, impact chronology, and mission planning</b>. Its emphasis on polar regions provides valuable insights for future exploration activities, particularly in areas of interest for resource prospecting.