Beyond the Data Imbalance: Employing the Heterogeneous Datasets for Vehicle Maneuver.

This page contains the drone dataset used in our paper

Paper │ Documentation │ Download │ Tutorials

Contents

1. Overview
2. Getting Started
3. Dataset Information

Overview

This is the drone dataset captured over several intersections in Daejeon, South Korea. The dataset contains two directories: raw contains the trajectory data for the tracked objects in the intersection, while processed contains the processed information from the raw data used to train the model in our paper.

Getting Started

Download

You can download the dataset from the releases page. Extract the zip files and make sure the dataset directory structure is as follows:

data_root
├── processed
│   ├── conversion
│   ├── link_idx
│   ├── maneuver_index
│   ├── nearest_outlet_state
│   ├── outlet_node_state
│   ├── total_traj
│   ├── plots
│   └── folder_tree
└── raw
    ├── background
    ├── landmark
    ├── mapSegmentation
    ├── recordingMeta
    ├── segmentation
    ├── tracks
    └── tracksMeta

Examples

Please take a look at the example notebook on how to use the information contained in the dataset.

Dataset Information

Statistics

• No. of tracked objects at each intersection

Intersection Id	Cars	Pedestrian	Bicycle
1	788	135	46
3	71	0	1
4	184	0	5
5	110	105	8
6	77	90	4

• No. of trajectories for each class

Cars	Pedestrian	Bicycle
10,555	4,353	571

$*$ Trajectory: 5 second long for 10hz interval.

Cite

If you find this drone dataset or our paper helpful for your own research, please consider citing:

@inproceedings{Beyond2024Jeon,
  title={Beyond the Data Imbalance: Employing the Heterogeneous Datasets for Vehicle Maneuver Prediction},
  author={Hyeongseok Jeon, Sanmin Kim, Abi Rahman Syamil, Junsoo Kim and Dongsuk Kum},
  booktitle={Proceedings of the European Conference on Computer Vision},
  year={2024}
}