講演情報
[P2-24]Ensemble-Based Spatially Adaptive Building Extraction Taming Thailand's Diverse Land Use Types
*Bhanu Prasad Chintakindi1, Shenglong Chen2, Yoshiki Ogawa2, Yoshihide Sekimoto2 (1. Department of Civil Engineering, The University of Tokyo, Japan, 2. Center for Spatial Information Science, The University of Tokyo, Japan)
キーワード：
Ensemble Modeling、YOLOv8、Instance Segmentation、Land Use Types
Accurately extracting buildings across diverse regions in Thailand poses significant challenges due to varied building patterns and environmental complexities. This study introduces an innovative ensemble modeling approach tailored to specific land use types within Thailand's distinct regions. By ensembling individually tailored building extraction models trained on diverse land-use types, the proposed method enhances accuracy and adaptability. Key contributions include advanced ensemble techniques, comprehensive cross-validation, and benchmarking against established models. This approach not only improves generalization capability but also provides a robust solution for accurate building extraction, promising advancements in urban planning, resource management, and environmental monitoring applications across Thailand.