Skip to content

Seed guided neural metric learning approach for calculating trajectory similarities

Notifications You must be signed in to change notification settings

yaodi833/NeuTraj

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NeuTraj

This is a seed guided neural metric learning approach for calculating trajectory similarities.

Require Packages

Pytorch, Numpy, trajectory_distance

Running Procedures

Create Folders

Please create 3 empty folders:

*data: Path of the original data which is organized to a trajectory list. Each trajectory in it is a list of coordinate tuples (lon, lat).

*features: This folder contains the features that generated after the preprocessing.py. It contains four files: coor_seq, grid_seq, index_seq and seed_distance.

*model: It is used for placing the NeuTraj model of each training epoch.

Download Data

Due to the file limit of Github, we put the dataset on other sites. Please first download the data and put it in data folder. The toy dataset can be download at: https://www.dropbox.com/s/ejoo1j21vjq7t7a/toy_trajs?dl=0

Preprocessing

Run preprocessing.py. It filters the original data and maps the coordinates to grids. After such process, intermediate files which contain coor_seq, grid_seq, and index_seq are generated. Then, we calculate the pair-wise distance under the distance measure and get the seed_distance.

Training & Evaluating

Run train.py. It trains NeuTraj under the supervision of seed distance. The parameters of NeuTraj can be modified in /tools/config.py

About

Seed guided neural metric learning approach for calculating trajectory similarities

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages