Automated crater detection with human level performance

Crater cataloging is an important yet time-consuming part of geological mapping. We present automated Detection Algorithm (CDA) that competitive with expert-human researchers and hundreds times faster. The CDA uses multiple neural networks to process digital terrain model thermal infra-red imagery identify locate craters across the surface Mars. use additional post-processing filters refine remove potential false crater detections, improving our precision recall by 10% compared Lee (2019). now find 80% known above 3km in diameter, 7,000 potentially new (13% identified craters). median differences between catalog other independent catalogs 2-4% location in-line inter-catalog comparisons. has been used global maps for Mars, software generated are available at https://doi.org/10.5683/SP2/CFUNII.