Skip to content

PattonYu/fufi-didi-xian

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 

Repository files navigation

DiDi-Xi'an dataset for FUFI problem

paper

Dataset of paper: Probabilistic Fine-Grained Urban Flow Inference with Normalizing Flows. If you find our paper or dataset useful for your research, please cite:

@inproceedings{Zhong2022FUFI,
 author = {Zhong, Ting and Yu, Haoyang and Li, Rongfan and Xu, Xovee and Luo, Xucheng and Zhou, Fan},
 title = {Probabilistic Fine-Grained Urban Flow Inference with Normalizing Flows},
 booktitle = {{ICASSP}},
 year = {2022}
} 
Data source

This dataset is collected by DiDi Chuxing GAIA Initiative, it contains about 1 Bilion trajectories of DiDi drivers in Xi’an, China from October 1 to October 31, 2016.

Data processing

In this dataset, we divide data into training, validation and test set in a ratio of 2 : 1 : 1. The dataset consists of twoparts, coarse- and fine-grained flow maps, where the coarse- maps are divided into pixel regions of size 32×32 and the fine- maps are divided into size of 128x128.

We collect meteorology factors of Xi'an from Wold Climate Data and the weather condition includes 9 categories (e.g., rainy, sunny and cloudy), then we digitize these categories into ordinal values.

We also include humidity and atmospheric pressure data for each flow map. Temperature, wind speed, humidity and pressure are scaled into the range [0, 1] with a min-max linear normalization.

Time and date factors (e.g., hour of the day, day of the week) are transformed into ordinal values.

Dataset description
Dataset DiDi-Xi’an
Time range 10/1/2016-10/31/2016
Time interval 10 minutes
Coarse-grained size 32 × 32
Fine-grained size 128 × 128
Upscaling factor (N) 4
Latitude range 34.20°N - 34.28°N
Longitude range 108.92°E - 109.01°E
Temperature/◦C [11.0, 32.0]
Wind speed / mph [0, 9.3]
Weather conditions 9 types
Humidity [36%, 100%]
Pressure /hPa [1004, 1026]
Holidays 7
Usage

The structure and usage details of the dataset can be refered to UrbanFM.

Unzip train.zip, test.zip, valid.zip into didixian folder, and then move didixian folder to dataset folder of UrbanFM, run the following command:

 python -m UrbanFM.train --ext_flag --dataset "didixian"

About

DIDI-Xi'an dataset for FUFI problem.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published