<i>Popnet</i>: computer vision based bespoken deep learning model for forecasting gridded population

This study introduces <i>Popnet</i>, a deep learning model for forecasting 1km-gridded populations, integrating U-Net, ConvLSTM, a Spatial Autocorrelation module and deep ensemble methods. Using spatial variables and population data from 2000 to 2020, <i>Popnet</i> predicts South Korea's population trends by age groups (under 14, 15-64, over 65) up to 2040. In validation, it outperforms traditional machine learning and state-of-the-art computer vision models. The output of this model discovered significant polarization: population growth in urban areas, especially the capital region, and severe depopulation in rural areas. <i>Popnet</i> is a robust tool for offering significant insights to policymakers and related stakeholders about the detailed future population, which allows them to establish detailed, localised planning and resource allocations.<br><br>*<i>Due to the export restrictions on grid data imposed by the National Geographic Information Institute of Korea, the training data has been replaced with data from Tennessee. However, the Korean version of the future prediction data remains unchanged. Please take this into consideration.</i><br>