Objective: Semantic segmentation of different classes in satellite imagery.
Models: Built U-net and Seg-Net models to generate masks.
Results: Using U-net model received Jaccard score 0.3752 on test data and for the Seg-Net test Jaccard score as 0.4463.
Use-cases: The model can be trained for various image segmentation tasks in satellite imagery, can be used in defense and security systems, creating and keeping up-to-date maps, disaster relief, and improving urban planning.