Satellite Imagery

TopCoder Marathon Match / Data Science 2nd Place The US National Geospatial-Intelligence Agency (NGA) was seeking novel approaches for satellite imagery segmentation in order to detect, delineate, and describe circular shaped features. My final solution reached 92% on their secret dataset. NGA's announcement (PyTorch, OpenCV, GDAL)

CVPR (Workshop 2019) Automatic system for 3D building reconstruction for the IARPA CORE3D program. This works received the Best Paper Award.