Efficacy and Efficiency in Object Detection: Exploring YOLOv9 on Limited Resources

Yangkai Zhou

2024

Abstract

The introduction of the You Look Only Once v9 (YOLOv9) algorithm marks a significant milestone in the realm of object detection, notably tackling the inherent information bottleneck in deep learning while enhancing model accuracy and parameter efficiency across various tasks. This advancement translates into expedited and more precise detection capabilities, surpassing its predecessor, the YOLO algorithm, as well as other modelling methodologies. This paper endeavours to delve into the performance evaluation of YOLOv9 and explore strategies for effectively training the model on small datasets and with limited computational resources. Through meticulous hyperparameter tuning, careful selection of optimizers, and comparison between YOLOv9-c and YOLOv9-e model variants, this study aims to ascertain the most suitable learning rates, batch sizes, and optimizers to optimize training efficacy. The insights garnered from this research serve as a valuable guide for small research teams and individuals facing computational constraints, providing them with a robust framework to streamline experimental processes and enhance overall efficiency in model development and training procedures. Ultimately, this paper contributes to advancing the accessibility and effectiveness of object detection methodologies in diverse settings and applications.

Paper Citation

in Harvard Style

Zhou Y. (2024). Efficacy and Efficiency in Object Detection: Exploring YOLOv9 on Limited Resources. In Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI; ISBN 978-989-758-713-9, SciTePress, pages 390-398. DOI: 10.5220/0012938700004508

in Bibtex Style

@conference{emiti24,
author={Yangkai Zhou},
title={Efficacy and Efficiency in Object Detection: Exploring YOLOv9 on Limited Resources},
booktitle={Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI},
year={2024},
pages={390-398},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012938700004508},
isbn={978-989-758-713-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI
TI - Efficacy and Efficiency in Object Detection: Exploring YOLOv9 on Limited Resources
SN - 978-989-758-713-9
AU - Zhou Y.
PY - 2024
SP - 390
EP - 398
DO - 10.5220/0012938700004508
PB - SciTePress