IIIT Hyderabad Publications |
|||||||||
|
Improved Road Connectivity by Joint Learning of Orientation and SegmentationAuthors: Anil Batra,C V Jawahar Conference: Computer Vision and Pattern Recognition 2019 (CPVR 2019) Date: 2019-06-16 Report no: IIIT/TR/2019/62 AbstractRoad network extraction from satellite images often pro-duce fragmented road segments leading to road maps unfit for real applications. Pixel-wise classification fails to pre-dict topologically correct and connected road masks due to the absence of connectivity supervision and difficulty in en-forcing topological constraints. In this paper, we propose a connectivity task called Orientation Learning, motivated by the human behavior of annotating roads by tracing it at a specific orientation. We also develop a stacked multi- branch convolutional module to effectively utilize the mu- tual information between orientation learning and segmen- tation tasks. These contributions ensure that the model pre- dicts topologically correct and connected road masks. We also propose Connectivity Refinement approach to further enhance the estimated road networks. The refinement model is pre-trained to connect and refine the corrupted ground- truth masks and later fine-tuned to enhance the predicted road masks. We demonstrate the advantages of our ap- proach on two diverse road extraction datasets SpaceNet [30] and DeepGlobe [11]. Our approach improves over the state-of-the-art techniques by 9% and 7.5% in road topol- ogy metric on SpaceNet and DeepGlobe, respectively. Full paper: pdf Centre for Visual Information Technology |
||||||||
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved. |