ChangeCoTNet : Building Change Detection by Contextual Transformer Deep Network

Document Type : Original Article

Authors

1 School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran

2 School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran

3 Dept. Of photogrammetry, School of Surveying and Geo Spatial Engineering, University of Tehran

10.22059/eoge.2025.378306.1153

Abstract

Building change detection (BCD) is a critical task in remote sensing, with applications in urban management and disaster assessment. However, achieving high accuracy in high-resolution BCD remains challenging due to the complexity of urban scenes. In this study, we propose ChangeCoTNet, a novel dual-branch deep learning model that integrates Contextual Transformer (CoT) blocks in the encoder and a Convolutional Neural Network (CNN) in the decoder. The CoT blocks enable the extraction of both static and dynamic contextual representations, while the Channel Attention Block (CAB) enhances discriminative feature extraction. The proposed model was implemented and evaluated on the LEVIR-CD and 2DCD datasets using a PyTorch backend. Experimental results demonstrate that ChangeCoTNet outperforms state-of-the-art methods, achieving F1-score improvements of 1.1% and 1.9% for the respective datasets. These results validate the effectiveness and efficiency of the proposed model in detecting changes with high precision and recall, making it a valuable tool for real-world applications.

Keywords

Main Subjects