You can install the required packages in your python environment by:
python>=3.7
torch>=1.7
pandas
numpy
matplotlib
pathlib
shutil
datetime
colored
math
Download the data with permission of using your own GPS trajectory data
Change the corresponding paths in the config.py and main.py files
Run and train the model by:
python main.py
The code will save the trained model every 10 epoches
Refer to Traj_Generation.ipynb file, we provide the fine-trained model (model.pt) simple guide infromation(heads.npy) example to explain how to generate trajectories.
If you use this code, please cite:
@inproceedings{zhu2023DiffTraj,
author = {Yuanshao Zhu, Yongchao Ye, Shiyao Zhang, Xiangyu Zhao and James, J.Q. Yu},
title = {DiffTraj: Generating GPS Trajectory with Diffusion Probabilistic Model},
booktitle = {Proceedings of the 37th Annual Conference on Neural Information Processing Systems},
year = {2023}
}