PEER-REVIEWED · AMERICAN METEOROLOGICAL SOCIETY
Peer-reviewed in the AMS journal Artificial Intelligence for the Earth Systems (2026)
“Operational Anomaly Detection in Satellite Imagery Using ResNet-18 Transfer Learning”
David Daniel, Aeroxis co-founder and CTO, is second author. The work was part of NOAA's Earth Observation Digital Twin (EODT) project. Aeroxis Enterprises delivered this work as a subcontractor to Science & Technology Corporation (STC), the prime contractor, with David Daniel as Aeroxis CTO.
“…our ResNet-18 model was able to detect anomalies in ABI with 98% accuracy…” (Sec. 3a)
“…the VIIRS model was able to detect synthetic striping anomalies with 95% accuracy…” (Sec. 3b)
Artificial Intelligence for the Earth Systems · Vol. 5, No. 1 · doi:10.1175/AIES-D-25-0026.1