TERRA-CD: A Multi-Temporal Benchmark Dataset for Urban Multi-Class and Semantic Change Detection

Omkar Oak$^{1^*}$, Rukmini Nazre$^1$, Rujuta Budke$^1$, Suraj Sawant $^1$
$^1$ COEP Technological University, Pune, India.

Abstract

Urban vegetation monitoring plays a vital role in understanding environmental changes and pro moting sustainable development, 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 paired Sentinel-2 images from 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 land cover transitions. Using various deep learning approaches including Siamese networks, STANet variants, Post-Classification Comparison, and HRSCD strategies, we evaluated the dataset’s effectiveness for both vegetation Multi-class Change Detection (MCD) as well as Semantic Change Detection (SCD). Our experiments achieved notable results, with these models averaging 97.5% accuracy and 65.7% kappa score for vegetation MCD, while SCD results averaged 93.2% accuracy and 57.1% kappa score. These results demonstrate TERRA-CD’s value as a versatile benchmark dataset for developing and evaluating urban vegetation monitoring systems and broader land cover change detection applications.

Citation

If you use this code or the findings in your research, please cite our paper:

Coming Soon

Contact

For any questions or inquiries about this research, please open an issue on this repository or contact the corresponding author.