A Local-Global Approach to Semantic Segmentation in Aerial Images

Aerial images are often taken under poor lighting conditions and contain low resolution objects, many times occluded by other objects. In this domain, visual context could be of great help, but there are still very few papers that consider context in aerial image understanding and still remains an open problem in computer vision. We propose a dual-stream deep neural network that processes information along two independent pathways. Our model learns to combine local and global appearance in a complementary way, such that together form a powerful classifier. We test our dual-stream network on the task of buildings segmentation in aerial images and obtain state-of-the-art results on the Massachusetts Buildings Dataset. We study the relative importance of local appearance versus the larger scene, as well as their performance in combination on three new buildings datasets. We clearly demonstrate the effectiveness of visual context in conjunction with deep neural networks for aerial image understanding.