DeepSUMO is a library that aims to connect the traffic simulator SUMO with current deep learning libraries like PyTorch to enable the use of data generated by SUMO with applications such as traffic prediction with graph neural networks.
I intend to further develop this framework (and readme) if I find the time to do so. An example of using the framework can be found in the example folder.
A quick thank you to Julie Wang, Amelia and Tracy Cai as their implementation of the ST-GAT model is used as an example here. The original code can be found here.