ANNA: A Deep Learning-Based Dataset in Heterogeneous Traffic for Autonomous Vehicles

Overview

This repository hosts the ANNA dataset, a comprehensive collection of images from Bangladeshi traffic, tailored for training and testing autonomous vehicle systems in heterogeneous traffic environments.

Dataset Description

• Link: The dataset is available in this link.
• Structure: Organized following the YOLOv5 folder structure.
• Size: Comprises 1800 images.
• Classes: Includes the following vehicle classes: cars, buses, rickshaws, bikes, CNGs, bicycles, easybikes, and vans.

Resources

• Trained Model: The trained model is available in this link.
• Demo Video: Demonstrates object detection on Bangladeshi roads.

License

This project is licensed under the MIT License.

Citation

If you use any part of the dataset in your work, please use the following BibTeX entries:

@misc{kamal2024anna,
      title={ANNA: A Deep Learning Based Dataset in Heterogeneous Traffic for Autonomous Vehicles}, 
      author={Mahedi Kamal and Tasnim Fariha and Afrina Kabir Zinia and Md. Abu Syed and Fahim Hasan Khan and Md. Mahbubur Rahman},
      year={2024},
      eprint={2401.11358},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

paper link

Nine types of objects are annotated in this dataset. Numbering of these 9 classes in the dataset are from 0 to 8 according to the order below (their images are added):

Human
Car
Bus
Rickshaw
Bike
CNG
Bicycle
Easybike
Van