High-Resolution Semantic Labeling with Convolutional Neural Networks

TL;DR

This paper analyzes existing CNN architectures for dense semantic labeling, identifies their limitations, and proposes a new CNN framework that effectively combines multi-resolution features, outperforming previous methods on aerial image benchmarks.

Contribution

The paper provides an in-depth analysis of semantic labeling CNNs and introduces a novel framework that better exploits local and global features for improved accuracy.

Findings

Outperforms previous techniques on aerial image benchmarks

Effectively combines multi-resolution features

Provides insights into properties of ideal semantic labeling CNNs

Abstract

Convolutional neural networks (CNNs) have received increasing attention over the last few years. They were initially conceived for image categorization, i.e., the problem of assigning a semantic label to an entire input image. In this paper we address the problem of dense semantic labeling, which consists in assigning a semantic label to every pixel in an image. Since this requires a high spatial accuracy to determine where labels are assigned, categorization CNNs, intended to be highly robust to local deformations, are not directly applicable. By adapting categorization networks, many semantic labeling CNNs have been recently proposed. Our first contribution is an in-depth analysis of these architectures. We establish the desired properties of an ideal semantic labeling CNN, and assess how those methods stand with regard to these properties. We observe that even though they provide competitive results, these CNNs often underexploit properties of semantic labeling that could lead to more effective and efficient architectures. Out of these observations, we then derive a CNN framework specifically adapted to the semantic labeling problem. In addition to learning features at different resolutions, it learns how to combine these features. By integrating local and global information in an efficient and flexible manner, it outperforms previous techniques. We evaluate the proposed framework and compare it with state-of-the-art architectures on public benchmarks of high-resolution aerial image labeling.