TT-CNN: a twin-tower convolutional spatial learned cost model for end-to-end runtime prediction

This package comprises: (1) Python scripts for a twin-tower convolution (TT-CNN)-based spatial learning model for end-to-end runtime prediction; (2) data files for evaluating the predictive performance of the TT-CNN model, including the training, validation, and test sets; and (3) Using large-scale OpenStreetMap (OSM) building-footprint polygons for California and New York, we construct standardized overlay datasets.