wellwang / deepist

Image-based NN model for travel time estimation for paths in road networks

DeepIST

DeepIST aims to predict the travel time of a given path (i.e., a sequence of road segments) in a road network. Please refer the paper here.

Prepare Data

If you don't want to parse data from scratch by yourself, you can skip this step and use our released data.

1. Map matching on trajectories

parse_[city].py <raw_fname> <output_traj_fname>
# for other cities: porto (well be released soon)

In this work, we apply barefoot for map matching based on open street map(OSM) data.

Some downloadable OSM data:
portugal
major cities

Scripts we implemented for barefoot will be released soon.

2. Parse map matching results as paths and filter incorrect matched results

python tools/path/traj_to_path.py <traj_file> <matched_folder> <output_path_file>
python tools/path/filter_paths.py <traj_file> <path_file> <output_path_file>

3. Plot paths as images

3.1. Prepare hourly moving speed on road segments in average based on path data

python tools/plot/get_road_avg_speed.py <path_file> <output_speed_file>

3.2. Plot paths as images

mkdir <output_image_folder>
python tools/plot.py <path_file> <speed_file> <osm.pbf_file> <output_image_folder> <output_training_file>

Released data

Porto

• raw trajectory data here
• trajectory data here
• path data here
• speed data here
• osm data here
• images here
• training file here

Chengdu

• Be released soon

How to Use?

First, to configurations of experiments in config.py
Then, to run DeepIST experiments, execute the following command:

python main.py <training_file>

Citing

If you find DeepIST useful for your research, please cite the following paper:

@inproceedings{fu2019deepist,
    title={DeepIST: Deep Image-based Spatio-Temporal Network for Travel Time Estimation},
    author={Fu, Tao-yang and Lee, Wang-Chien},
    booktitle={Proceedings of the 28th ACM International Conference on Information and Knowledge Management},
    pages={69--78},
    year={2019},
    organization={ACM}
}

Miscellaneous

Please send any questions you might have about the code and/or the algorithm to [email protected] or [email protected].