Package that provides neural networks (mostly autoencoders) to embed and generate traffic trajectories. This project relies on traffic and Pytorch-Lightning libraries.
# create new python environment for traffic
conda create -n traffic -c conda-forge python=3.9 traffic
conda activate traffic
# clone project
git clone https://github.com/kruuZHAW/deep-traffic-generation
# install project
cd deep-traffic-generation
pip install .
Navigate to any python file in deep_traffic_generation
and run it.
# module folder
cd deep_traffic_generation
# example: run module with custom arguments
python tcvae.py --data_path *path to data* --prior vampprior --encoding_dim 32 --h_dims 64 64 64 --lr 0.001 --batch_size 100 --n_components 200 --features track groundspeed altitude timedelta --info_features latitude longitude --info_index -1
You can use Tensorboard to visualize training logs.
tensorboard --logdir lightning_logs