A Yolo-Based Deep Learning Framework for Accurate Multi-Object Counting in Complex and Crowded Scenes

Sunil Kumar, K. Sindhuja, Kalpesh Rasiklal Rakholia, Lokasani Bhanuprakash, Bhavanath J.

2025

Abstract

Accurate multi-object counting in complex and crowded scenarios is still posing a critical challenge in computer vision, especially in the presence of occlusion, multi-object scales, and real-time considerations. We proposed a YOLO3205122*2 deep learning framework which is tailored for accurate counting and high-density objects detection. The proposed method handles the problem of overlapping objects and the diverse poses through the use of the contextual awareness, the and the attention model, being also able to benefit from low-latency inference. We further verify the model over aerial and ground level datasets, achieving state-of-the-art results in real surveillance and crowd analysis. Moreover, additional techniques of transformer-assisted decoding, deformable convolutions, and optimized deployment for edge devices contribute to further scalability and deployment flexibility. Extensive experiments demonstrate that our framework outperforms classical YOLO baselines in terms of accuracy, speed and generalization, and therefore constitutes a novel state-of-the-art framework for counting multiple objects in complex visual scenarios.

Paper Citation

in Harvard Style

Kumar S., Sindhuja K., Rakholia K., Bhanuprakash L. and J. B. (2025). A Yolo-Based Deep Learning Framework for Accurate Multi-Object Counting in Complex and Crowded Scenes. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - Volume 1: ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 396-401. DOI: 10.5220/0013866400004919

in Bibtex Style

@conference{icrdicct`2525,
author={Sunil Kumar and K. Sindhuja and Kalpesh Rakholia and Lokasani Bhanuprakash and Bhavanath J.},
title={A Yolo-Based Deep Learning Framework for Accurate Multi-Object Counting in Complex and Crowded Scenes},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - Volume 1: ICRDICCT`25},
year={2025},
pages={396-401},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013866400004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - Volume 1: ICRDICCT`25
TI - A Yolo-Based Deep Learning Framework for Accurate Multi-Object Counting in Complex and Crowded Scenes
SN - 978-989-758-777-1
AU - Kumar S.
AU - Sindhuja K.
AU - Rakholia K.
AU - Bhanuprakash L.
AU - J. B.
PY - 2025
SP - 396
EP - 401
DO - 10.5220/0013866400004919
PB - SciTePress