The View-of-Delft Prediction devkit

Welcome to the View-of-Delft Prediction (VoD-P) development kit. This repository contains the code and documentation associated with the VoD-P dataset.

Overview

• Introduction
• Changelog
• Devkit setup
• VoD-P
  • VoD-P setup
  • Getting started with VoD-P
  • Submitting to the VoD-P leaderboard
• Citation

Introduction

The View-of-Delft Prediction dataset is an extension of the View-of-Delft dataset. It contains the 3D object annotations of the original dataset and additionally provides accurate 6-DoF global localisation and semantic map data.

The dataset is available in a format based on the nuScenes dataset, and hence this development kit is a modified version of the nuScenes devkit.

Changelog

• [2024-11-15] Released a version of the development kit for Python 3.8.
• [2024-09-11] Released the View-of-Delft Prediction dataset and development kit.

Devkit setup

The devkit is tested for Python 3.8. For a version of the devkit that is compatible with Python 3.6 and 3.7, see the v1.0.1 PyPI release or tag. To install Python, please check here.

Our devkit is available and can be installed via pip:

pip install vod-devkit

For an advanced installation, see installation for detailed instructions.

VoD-P

VoD-P setup

To download VoD-P, follow the instructions at the main View-of-Delft dataset page. Download the zipfile when you receive the access link. Unzip the file and you should have the following folder structure:

/data/sets/vod
    maps	-	Folder for all map files (vectorized .json files).
    v1.0-*	-	JSON tables that include all the meta data and annotations. Each split (trainval, test) is provided in a separate folder.

Getting started with VoD-P

Please follow these steps to make yourself familiar with the VoD dataset:

• Read the main dataset page.
• Request access to the dataset.
• Download the dataset.
• Get the vod-devkit code.
• Read the tutorials or run one yourself using:

jupyter notebook $HOME/vod-devkit/tutorials/vod_tutorial.ipynb

• Read the View-of-Delft Prediction paper for a closer look at the dataset.
• See the FAQs.

Submitting to the VoD-P leaderboard

The VoD-P benchmark leaderboard can be found at TODO.

See the benchmark instructions for the submission format and rules.

Citation

Please use the following citation when referencing the View-of-Delft (VoD-P) dataset:

@article{boekema2024vodp,
  author={Boekema, Hidde J-H. and Martens, Bruno K.W. and Kooij, Julian F.P. and Gavrila, Dariu M.},
  journal={IEEE Robotics and Automation Letters}, 
  title={Multi-class Trajectory Prediction in Urban Traffic using the View-of-Delft Prediction Dataset}, 
  year={2024},
  volume={9},
  number={5},
  pages={4806-4813},
  keywords={Trajectory;Roads;Annotations;Semantics;Pedestrians;Predictive models;History;Datasets for Human Motion;Data Sets for Robot Learning;Deep Learning Methods},
  doi={10.1109/LRA.2024.3385693}}