Subventions et des contributions :

Subvention ou bourse octroyée s'appliquant à plus d'un exercice financier (2017-2018 à 2018-2019).

In collaboration with UrtheCast inc., the goal of this project is to explore the concrete usability of deepx000D
convolutional neural networks in the field of satellite imagery. Since this project spans over a short period ofx000D
time (only 6 months) we will focus on image segmentation, a topic at the core of the research activities of pr.x000D
Jodoin and a glaring issue for UrtheCast. As such, we look forward to answer the following four questions:x000D
1) Which state-of-the-art network configuration works best and what is its accuracy compared to thex000D
none-deep-learning solutions currently deployed by UrtheCast?x000D
2) How many manually annotated satellite images are required to properly train a convolutional neuralx000D
network? Also, does weakly annotated data can be used to increase the accuracy of a network?x000D
3) How can very large satellite images (from 2,000 x 2,000 to 6,000 x 6,000 pixels) be processed (both atx000D
training and at test time) on a single 12 Gb GPU?x000D
4) Can a model trained on a RGB dataset be transferred to a multispectral dataset as well as data from thex000D
UrtheCast satellites?x000D
Since deep segmentation models are new to the world of remote sensing, this project is fundamentallyx000D
important both for UrtheCast and the Canadian society as a whole. With its "UrtheDaily(TM)" project,x000D
UrtheCast is about to deploy a constellation of satellites which will acquire an unprecedented number ofx000D
images that no one can manually analyse. This project will thus provide the company with the tools forx000D
processing those images. The same models could also be used to process data acquire by other Canadianx000D
satellites such as RadarSat2 and SCISAT.