This is the companion code for the training method reported in Cheap and Deterministic Inference for Deep State-Space Models of Interacting Dynamical Systems by Andreas Look et al., TMLR 2023.
This repo contains a tutorial like jupyter notebook examples/multimodal_toy.ipynb to recreate the toy example from Fig. 2 in our paper as well
as the training script scripts/train.py for the traffic forecasting experiments. After training, predictions can be visualized with examples/visualize.ipynb.
The rounD dataset can be processed with the file data/preprocess_rounD.py. We filter out non-moving objects as well as pedestrians.
The NGSIM dataset can be processed with code from Nachiket Deo and Mohan M. Trivedi.
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
This project is open-sourced under the AGPL-3.0 license. See the LICENSE file for details.
For a list of other open source components included in this project, see the file 3rd-party-licenses.txt.