LW-YOLOv8: An Lightweight Object Detection Algorithm for UAV Aerial Imagery

Hu Chen*, Dianwei Wang, Jie Fang, Yuanqing Li, Sidi Xu, Zhijie Xu

*Corresponding author for this work

Research output: Chapter in Book or Report/Conference proceedingConference Proceedingpeer-review

1 Citation (Scopus)

Abstract

Deep learning is vital for improving drone intelligence through UAV aerial object detection. However, current algorithms struggle to balance detection speed and accuracy. To address this issue, in this paper we propose a lightweight UAV aerial image object detection algorithm based on YOLOv8 named LW-YOLOv8. Firstly, the lightweight VanillaNet architecture is used in the backbone network to effectively reduce the number of YOLOv8 network parameters. Then, a feature pyramid network AFPNs based on shallow network fusion is proposed to improve the detection performance of small targets in UAV aerial images. Experimental results on the VisDrone2019 dataset show that compared with the YOLOv8 algorithm, the proposed algorithm reduces the number of parameters and improves the detection accuracy simultaneously.

Original languageEnglish
Title of host publication2024 6th International Conference on Natural Language Processing, ICNLP 2024
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages446-450
Number of pages5
ISBN (Electronic)9798350349115
DOIs
Publication statusPublished - 2024
Externally publishedYes
Event6th International Conference on Natural Language Processing, ICNLP 2024 - Hybrid, Xi'an, China
Duration: 22 Mar 202424 Mar 2024

Publication series

Name2024 6th International Conference on Natural Language Processing, ICNLP 2024

Conference

Conference6th International Conference on Natural Language Processing, ICNLP 2024
Country/TerritoryChina
CityHybrid, Xi'an
Period22/03/2424/03/24

Keywords

  • lightweight YOLOv8
  • object detection
  • UAV aerial images

Fingerprint

Dive into the research topics of 'LW-YOLOv8: An Lightweight Object Detection Algorithm for UAV Aerial Imagery'. Together they form a unique fingerprint.

Cite this