Seattle Demo Accompanying Files

We release the data, code, and prepared city graph objects to facilitate city scale building operating energy prediction with Seattle as a case study.

The zipped folder consists of five separate folders:

• Trained_Model (Pretrained GNN model weights in PyTorch format)
• Seattle Graphs (Contains city object nodes and COO format edges)
• LCZ_raster (100M global local climate zone raster, obtained from https://doi.org/10.5194/essd-14-3835-2022)
• Image_Model (Pretrained Resnet-18 image encoder)
• Building_Satellite (Placeholder folder for building satellite images)

Due to file storage limitations, building satellite files are separately available at: 1) 10.6084/m9.figshare.28188230, 2)10.6084/m9.figshare.28188005, and 3) 10.6084/m9.figshare.27091783.

The provided files are supplementary to the code repository which provides Python notebooks stepping through the data preprocessing, GNN training, and satellite imagery download processes.

For any questions or clarifications, please contact: winyap@mit.edu.