Fully Convolutional Network for Automatic Road Extraction From Satellite Imagery

Alexander Buslaev, Selim Seferbekov, Vladimir Iglovikov, Alexey Shvets; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 207-210

Abstract


Analysis of high resolution satellite images has been an important research topic for traffic management, city planning and road monitoring. One of the problem here is automatic and precise road extraction. From an original image it is difficult and computationally expensive to extract roads due to presences of other road-like features with straight edges. In this paper we propose an approach for automatic road extraction based on fully convolutional neural network of U-net family. This network is consisted of ResNet-34 pre-trained on ImageNet and decoder adapted from vanilla U-Net. Based on validation results, leaderboard and our own experience this network shows superior results for the DEEPGLOBE - CVPR 2018 road extraction sub-challenge. Moreover, this network uses moderate memory that allows to use just one GTX 1080 or 1080ti video cards to preform whole training and makes pretty fast predictions.

Related Material


[pdf]
[bibtex]
@InProceedings{Buslaev_2018_CVPR_Workshops,
author = {Buslaev, Alexander and Seferbekov, Selim and Iglovikov, Vladimir and Shvets, Alexey},
title = {Fully Convolutional Network for Automatic Road Extraction From Satellite Imagery},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2018}
}