Citation

Lim, Chin Hong and Tee, Connie and Goh, Michael Kah Ong (2024) Vehicle Detection Based on Improved YOLOv8. In: 2024 IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET), 26-28 August 2024, Kota Kinabalu, Malaysia.

Text Vehicle Detection Based on Improved YOLOv8.pdf - Published Version Restricted to Repository staff only Download (330kB)

Abstract

Vehicle detection is a crucial component of Intelligent Transportation System (ITS). With the advancement of deep learning and computer vision, state-ofthe-art algorithms such as You Only Look Once version 8 (YOLOv8) have been developed to address the real-time detection challenges. However, during the experimentation process, we observed a class imbalance problem in the selfcollected Malaysian traffic dataset. In this paper, we analyse this problem in detail and propose a Simplified Variant of Oversampling (SVO) method to mitigate it. By applying the proposed method, we can ensure that the detection algorithm performs consistently across all classes, thereby improving the overall accuracy and dependability of the system. The measurements used in this paper are mean Average Precision (mAP), precision and recall. The oversampling approach improved mAP50 and mAP50-95 by 16.91% and 29.43%, respectively. Additionally, different attention mechanisms for instance Triplet Attention, Squeeze and Excitation Networks (SE) and Large Selective Kernel Network (LSKNet), were integrated into the backbone of the detection network to further enhance performance. Among these, the LSKNet attention mechanism achieved the best performance, further improving the mAP50 and mAP50-95 by 2.46% and 2.70%, respectively.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Vehicle Detection, Computer Vision
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Nurul Iqtiani Ahmad
Date Deposited:	05 Dec 2024 02:43
Last Modified:	05 Dec 2024 02:43
URII:	http://shdl.mmu.edu.my/id/eprint/13240

Downloads

Edit (login required)