Deep multi-feature learning architecture for water body segmentation from satellite images

Rishikesh G. Tambe, Sanjay N. Talbar, Satishkumar S. Chavan

Abstract: Highlights•Water body segmentation is crucial for many applications.•W-Net consists of inception blocks at both encoder and decoder paths.•Use of asymmetric convolution reduces trainable parameters within the network.•Refinement modules enhances predicted image and identifies boundary pixels accurately.