A2D2-AUDI-Dataset Computer Vision Project

Original source of the dataset can be found in the following link: https://www.a2d2.audi/a2d2/en/download.html

2D Bounding boxes information were extracted from 3D bounding box annotations from the following link. These annotations then were converted to YOLO format. Original images were resized from 1920x1280 px to 416x416 px to be able to use in local computers with no-GPU/1-GPU. Then randomly selected 3000 images divided into 2000-500-500 train-validation-test folders and uploaded to Roboflow.

Full dataset can be downloaded from the following link. Dataset – 3D Bounding Boxes README - 3D Boxes

Paper: Pdf, Arxiv

Please cite as follows for the original source of the dataset:

@article{geyer2020a2d2,

title={{A2D2: Audi Autonomous Driving Dataset}},

author={Jakob Geyer and Yohannes Kassahun and Mentar Mahmudi and Xavier Ricou and Rupesh Durgesh and Andrew S. Chung and Lorenz Hauswald and Viet Hoang Pham and Maximilian M{"u}hlegg and Sebastian Dorn and Tiffany Fernandez and Martin J{"a}nicke and Sudesh Mirashi and Chiragkumar Savani and Martin Sturm and Oleksandr Vorobiov and Martin Oelker and Sebastian Garreis and Peter Schuberth},

year={2020},

eprint={2004.06320},

archivePrefix={arXiv},

primaryClass={cs.CV},

url = {https://www.a2d2.audi} }

Prepared by: Sancak Özdemir