Detecting Abandoned Houses in Rural Areas using Multi-Source Data

Abstract Abandoned houses have become a common feature of the local landscapes: rising number abandoned is major challenge facing many counties in South Korea. Their presence negatively influences neighborhood by undermining its aesthetic quality, depreciating perception safety properties, and deepening fiscal deficit financing. The detection first step toward adequate housing management governments. This study aims to provide cost-effective prompt approach identifying rural areas. Multi-source data, that is, images building registry data are utilized multi-input neural network designed adopt these heterogeneous datasets. Trained two source datasets, proposed achieves 86.2% accuracy classifying houses, which an acceptable performance level administrative practice. database identified this manner expected promote effective governments ultimately contribute mitigating vacancies