Urban Change Detection Using a Dual-Task Siamese Network and Semi-Supervised Learning

TL;DR

This paper introduces a semi-supervised dual-task Siamese network for urban change detection that leverages unlabeled data to improve accuracy, demonstrating superior performance on the SpaceNet7 dataset.

Contribution

It proposes a novel semi-supervised learning approach combined with a dual-task Siamese network architecture for urban change detection, enhancing results over fully supervised methods.

Findings

SSL improves change detection accuracy.

The dual-task network effectively segments buildings and detects changes.

Method outperforms baseline supervised models.

Abstract

In this study, a Semi-Supervised Learning (SSL) method for improving urban change detection from bi-temporal image pairs was presented. The proposed method adapted a Dual-Task Siamese Difference network that not only predicts changes with the difference decoder, but also segments buildings for both images with a semantics decoder. First, the architecture was modified to produce a second change prediction derived from the semantics predictions. Second, SSL was adopted to improve supervised change detection. For unlabeled data, we introduced a loss that encourages the network to predict consistent changes across the two change outputs. The proposed method was tested on urban change detection using the SpaceNet7 dataset. SSL achieved improved results compared to three fully supervised benchmarks.