Dense semantic labeling of sub-decimeter resolution images with convolutional neural networks

TL;DR

This paper introduces a CNN-based system for dense semantic labeling of ultra-high resolution aerial images, achieving state-of-the-art accuracy, improved geometric precision, and high efficiency by using a downsample-then-upsample architecture.

Contribution

The paper presents a novel CNN architecture with a downsample-then-upsample design for pixel-level land-cover classification in ultra-high resolution imagery, outperforming existing methods.

Findings

Achieves state-of-the-art accuracy on sub-decimeter datasets

Improves geometric accuracy of semantic labels

Offers high inference efficiency

Abstract

Semantic labeling (or pixel-level land-cover classification) in ultra-high resolution imagery (< 10cm) requires statistical models able to learn high level concepts from spatial data, with large appearance variations. Convolutional Neural Networks (CNNs) achieve this goal by learning discriminatively a hierarchy of representations of increasing abstraction. In this paper we present a CNN-based system relying on an downsample-then-upsample architecture. Specifically, it first learns a rough spatial map of high-level representations by means of convolutions and then learns to upsample them back to the original resolution by deconvolutions. By doing so, the CNN learns to densely label every pixel at the original resolution of the image. This results in many advantages, including i) state-of-the-art numerical accuracy, ii) improved geometric accuracy of predictions and iii) high efficiency at inference time. We test the proposed system on the Vaihingen and Potsdam sub-decimeter resolution datasets, involving semantic labeling of aerial images of 9cm and 5cm resolution, respectively. These datasets are composed by many large and fully annotated tiles allowing an unbiased evaluation of models making use of spatial information. We do so by comparing two standard CNN architectures to the proposed one: standard patch classification, prediction of local label patches by employing only convolutions and full patch labeling by employing deconvolutions. All the systems compare favorably or outperform a state-of-the-art baseline relying on superpixels and powerful appearance descriptors. The proposed full patch labeling CNN outperforms these models by a large margin, also showing a very appealing inference time.