datavisit/kdd_submission

This repository contains the code and visualization demo for our paper "Diff-ExpertNet: A Deep Expert Framework Using Conditional Diffusion Models for Vessel Trajectory Generation" submitted to [KDD] 2025.

Repository Structure

• code/: Contains the source code for this paper.
  • data_preprocessing/: Scripts for AIS data preprocessing.
  • models/: Implementation models of our proposed framework.
  • utils/: Utility scripts.
• map/: Map image, ports and coastlines.
• demo/: Visualization demo about experimental results.

Dataset

1. Download public AIS dataset from source.
2. Details about AIS data preprocessing can be found in code/data_preprocessing directory.

Visualization demo about our proposed vessel trajectory generation methods

We merged the vessel trajectories from the test set for January 2023 into a single day, and then provided a comparison between the ground-truth and generated results.