ORIGINAL RESEARCH article
Front. Plant Sci.
Sec. Sustainable and Intelligent Phytoprotection
This article is part of the Research TopicHighlights of 1st International Conference on Sustainable and Intelligent Phytoprotection (ICSIP 2025)View all 12 articles
A Lightweight Intelligent Grading Method for Lychee Anthracnose Based on Improved YOLOv12
Provisionally accepted
- Guangdong University of Petrochemical Technology, Maoming, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Abstract—Anthracnose is one of the primary diseases 5 leading to quality deterioration in lychee. Traditional 6 manual grading methods suffer from low efficiency 7 and high subjectivity. To achieve rapid, 8 non-destructive detection and intelligent grading of 9 lychee anthracnose, while addressing the challenge of 10 balancing high accuracy and lightweight design in 11 detection models, this study proposes a lightweight 12 improved model named LycheeGuard-Lite based on 13 the YOLOv12 framework. By introducing the 14 C3k2_Light module reconstructed with depthwise 15 separable convolutions, a dual-path C2PSA attention 16 mechanism (position-channel dual-path attention), and 17 the wConv2D weighted convolution strategy, the 18 model enhances lesion feature extraction capability 19 while reducing computational complexity.Evaluation 20 was performed on a self-built dataset comprising 21 14,576 images of two dominant lychee varieties 22 ('Feizixiao' and 'Baitangying') collected under multiple 23 lighting conditions and annotated with three severity 24 levels (Mild, Moderate, Severe). The results 25 demonstrate that the model maintains 99.4% mAP50 26 detection accuracy while reducing its number of 27 parameters to 2.19M (a 12.8% decrease) and 28 computational cost to 4.1 GFLOPs (a 29.3% 29 reduction).This research provides a lightweight and 30 deployable algorithmic foundation for automated 31 lychee disease recognition and intelligent grading, 32 offering practical engineering value for post-harvest 33 fruit sorting and quality management.
Keywords: lychee anthracnose, Disease classification, YOLOv12, Lightweight model, attention mechanism
Received: 19 Aug 2025; Accepted: 25 Nov 2025.
Copyright: © 2025 Xu, Ma, Su, He and Wu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Xianjun Wu
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.