City-scale Vehicle Trajectory Data From Traffic Camera Videos

Code of our dataset description paper:

Fudan Yu, Huan Yan, Rui Chen, Guozhen Zhang, Meng Chen and Yong Li. City-scale Vehicle Trajectory Data From Traffic Camera Videos (Scientific Data 2023).

Figure 1. Overall Framework.

Requirements

Python 3.7
numpy == 1.21.3
faiss == 1.5.3
coloredlogs == 15.0.1
osmnx == 1.1.2
networkx == 2.6.3
shapely == 1.8.0
matplotlib == 3.4.3
pyyaml == 6.0
pandas == 1.3.4
coord-convert == 0.2.1
pytorch == 1.10.1

These dependencies can be installed using the following command:

pip install -r requirements.txt

Another requirement (Fast Map Matching for py3) can be installed as follows:

Install:

apt install libboost-dev libboost-serialization-dev gdal-bin libgdal-dev make cmake libbz2-dev libexpat1-dev swig python-dev
mkdir build && cd build && cmake .. -DCMAKE_INSTALL_PREFIX=~/.local && make -j32 && make install

Test:

cd ../example/python && python fmm test.py

File Description

code

All the .py and .ipynb codes for how we generate the trajectory dataset based on visual embedded traffic camera records, evaluate the vehicle Re-ID and trajectory recovery metrics, and report statistical characteristic.

For details, see the README in this directory.

dataset

The .csv files as the proposed dataset. Please download from our Figshare repository.

example

A .ipynb example on the basic usage of the proposed dataset.

Citation

If you find this repository useful in your research, please consider citing the following paper:

@Article{Yu2023,
author={Yu, Fudan and Yan, Huan and Chen, Rui and Zhang, Guozhen and Liu, Yu and Chen, Meng and Li, Yong},
title={City-scale Vehicle Trajectory Data from Traffic Camera Videos},
journal={Scientific Data},
year={2023},
month={Oct},
day={17},
volume={10},
number={1},
pages={711},
issn={2052-4463},
doi={10.1038/s41597-023-02589-y},
url={https://doi.org/10.1038/s41597-023-02589-y}
}