Deep Learning at SARLab

Deep learning has tremendous potential to leverage the massive quantities of data available from remote sensing platforms. At SARLab, we have access to terrabytes of RADARSAT-2 synthetic aperture radar data, which provides an image resolution of up to 3M per pixel. This level of detail allows for tasks such as road extraction, ship detection and land use classification, but the full potential for exploration of this data with deep learning has only just begun.

Diverse and challenging research opportunities. Synthetic aperture radar is a highly complex image processing domain with many challenging research directions. Whether your interest is new methods of phase analysis or strategies for state-of-the-art environmental monitoring, from computer vision to time series analysis, SARLab has a place for you to grow as a deep learning researcher.

State of the art computational resources. We'd all rather spend less time waiting for our models to train. At SARLab, you'll have access to to two GPU server racks: SARSERV-1 which has 4 dual-GPU NVIDIA Tesla K80s, and SARSERV-3 which has 4 NVIDIA Tesla V100s (32GB).