Context-Enhanced Multi-View Trajectory Representation Learning: Bridging the Gap through Self-Supervised Models

This is a PyTorch implementation of SMARTraj: A Stable Multi-City Adaptive Method for Multi-View Spatio-Temporal Trajectory Representation Learning SMARTraj for trajectory representation learning as described in our paper

Requirements

Our code is based on Python version 3.7 and PyTorch version 1.8.1. And other dependencies are given in requirement.txt.

data

We conduct our experiments on two trajectory datasets: Xi'an and Chengdu. These datasets are the same as those used in our previous work,(MVTraj).You can access the data here.

Code Structure

You can find our pretraining code in model_train.py, and the model architecture is defined in SMARTraj.py. Before running the pretraining process, ensure that the xx.json configuration files in the config/ folder are correctly set. After pretraining, you can evaluate the model using the tasks provided in the downstream/ folder.