I work on oriented object detection for Earth Observation: ships, aircraft, vehicles, and infrastructure in satellite imagery. This blog documents that work—especially oriented-det, a sovereign, Apache-licensed detection stack for operational EO systems.
I founded DL4EO, where we help teams build and deploy deep learning for geospatial applications. Before that, I spent many years at Airbus Defence and Space Intelligence working on machine learning for satellite imagery, including public datasets and Kaggle competitions.
Older posts on this site cover axis-aligned detectors (YOLO family) on Airbus imagery; newer posts focus on oriented boxes, DOTA, and closing the gap with research frameworks like MMRotate.