Inter IIT Project

• Article: Learn How to Train U-Net On Your Dataset
• Repo: Previously implemented repo
• Paper: CNNs for large scale-remote sensing image Classification
• Paper: RNNs to correct Satellite Image Classification Maps
• Article: Deconvolution
• Report: Various indices from Satellite images

• Water = [0,0,150]
• Trees = [0,125,0]
• Grass = [0,255,0]
• Trains = [255,255,0]
• Soil = [150,80,0]
• Roads = [0,0,0]
• Unlabelled = [255,255,255]
• Buildings = [100,100,100]
• Pools = [150,150,255]

• road:15.2 0.5
• building:21.88 0.5
• tree:8.5 0.4
• grass:6.88 0.4
• soil:0.93 0.35
• water:5.72 0.4
• rail:1.0 0.35
• pool:0.15 0.3
• unlabelled:39.72 0.6