Abstract: Capturing sufficient global context and rich spatial structure information is critical for dense prediction tasks. Convolutional Neural Network (CNN) is particularly adept at modeling ...

Abstract: Geographic object segmentation is a fundamental yet challenging problem for remote sensing image interpretation. The prevalent paradigm to solve this problem is to train deep neural networks ...