TERRA-CD: Multi-Temporal Framework for Multi-class and Semantic Change Detection

TL;DR

TERRA-CD is a comprehensive benchmark dataset with diverse annotations and deep learning evaluations for multi-temporal urban vegetation and land cover change detection using Sentinel-2 imagery.

Contribution

It introduces a large, annotated dataset for multi-class and semantic change detection, along with evaluations of various deep learning methods.

Findings

Deep learning models effectively detect land cover changes.

Semantic change detection benefits from multi-class annotations.

The dataset enables benchmarking of change detection approaches.

Abstract

Urban vegetation monitoring plays a vital role in understanding environmental changes, yet comprehensive datasets for this purpose remain limited. To address this gap, we present the Temporal Remote-sensing Repository for Analyzing Change Detection (TERRA-CD), a benchmark dataset comprising 5,221 Sentinel-2 image pairs from 2019 and 2024, covering 232 cities across the USA and Europe. The dataset features three distinct annotation schemes: 4-class land cover mapping masks, 3-class vegetation change masks, and 13-class semantic change masks capturing all possible land cover transitions. Using various deep learning approaches including Siamese networks, STANet variants, Bi-SRNet, Changemask, Post-Classification Comparison, and HRSCD strategies, we evaluated the dataset's effectiveness for both vegetation Multi-class Change Detection as well as Semantic Change Detection. The proposed dataset and methods are available at https://github.com/omkarsoak/TERRA-CD.