YOLOv8-Scm: an improved model for citrus fruit sunburn identification and classification in complex natural scenes

Guo-xun Cong, Xinghong Chen, Zongyu Bing, Wenhuan Liu, Xiangling Chen, Yue Wang, Zheng Guo, Youliang ZhengFrontiers in Plant Science 2025 · DOI

Abstract

Citrus production faces significant economic threats from sunburn caused by extreme heat and solar radiation. Traditional identification methods often fail in complex orchard environments. This study introduces YOLOv8-Scm, a lightweight model based on YOLOv8n, optimized for real-time sunburn detection and grading. By incorporating Distribution-Shifted Convolution (DSConv), a Global Attention Mechanism (GAM), and the EIoU loss function, the model achieves a mAP50 of 92.7% and a Precision of 86.6%. These enhancements allow for superior performance in challenging natural scenarios with minimal computational overhead, providing a robust tool for smart orchard management.

Introduction

Citrus is a vital global crop, particularly in southern China. However, climate change has increased the frequency of "citrus sunburn," a physiological disorder that damages fruit skin and halts development. Automating the assessment of this damage is difficult due to environmental noise like variable lighting and leaf occlusion, the need for fine-grained severity grading, and the limited processing power of mobile devices used in orchards.

Methodology

The researchers developed YOLOv8-Scm by enhancing the YOLOv8n (nano) architecture with three specific components designed to balance efficiency and accuracy.

The model improvement workflow from baseline YOLOv8 to the enhanced YOLOv8-Scm.
  1. DSConv (Distribution-Shifted Convolution): Replaces standard convolutions in the backbone to reduce memory and computational overhead while maintaining feature quality.
  2. GAM (Global Attention Mechanism): Captures channel and spatial contextual cues to improve detection of occluded or multi-scale targets.
  3. EIoU (Enhanced Intersection over Union) Loss: Improves bounding box regression, leading to more accurate localization of fruits across various sizes.

Data Collection and Classification

The study utilized a custom dataset of 2,099 high-resolution images of Orah mandarins, captured under diverse conditions to ensure model robustness.

Environmental conditions and hardware parameters used for dataset collection.

The dataset includes 5,390 labeled instances categorized into a 5-level grading system:

  • Hfruits: Healthy (pure green).
  • ODB-fruit: Level 1 (slight yellowing).
  • SDB-fruit: Level 2 (discoloration < 25%).
  • TDB-fruit: Level 3 (burns covering 25%–66.7%).
  • FDB-fruit: Level 4 (severe burns/scabs > 66.7%).

Results and Discussion

YOLOv8-Scm demonstrates significant improvements over the baseline and other state-of-the-art models:

  • Performance: Achieved 92.7% mAP50, a 2.0% increase over YOLOv8n.
  • Efficiency: The model uses only 3.2M parameters, maintaining an inference speed of 117.6 FPS on a standard GPU.
  • Comparative Advantage: It outperformed SSD, Faster R-CNN, and YOLOv10n in terms of precision and recall for this specific agricultural task.

The inclusion of the Global Attention Mechanism was particularly effective in reducing false positives caused by complex backgrounds, such as mistaking green leaves for healthy fruit.

Conclusion

YOLOv8-Scm provides a highly efficient and accurate solution for monitoring citrus sunburn in real-time. Its lightweight design makes it ideal for deployment on mobile devices and drones, offering critical technical support for precision agriculture. Future work will focus on expanding the model's generalization to other citrus varieties and improving performance under extreme occlusion.

Summary updated

Abstract

Citrus ranks among the most widely cultivated and economically vital fruit crops globally, with southern China being a major production area. In recent years, global warming has intensified extreme weather events, such as prolonged high temperature and strong solar radiation, posing increasing risks to citrus production,leading to significant economic losses. Existing identification methods struggle with accuracy and generalization in complex environments, limiting their real-time application. This study presents an improved, lightweight citrus sunburn recognition model, YOLOv8-Scm, based on the YOLOv8n architecture. Three key enhancements are introduced: (1) DSConv module replaces the standard convolution for a more efficient and lightweight design, (2) Global Attention Mechanism (GAM) improves feature extraction for multi-scale and occluded targets, and (3) EIoU loss function enhances detection precision and generalization. The YOLOv8-Scm model achieves improvements of 2.0% in mAP50 and 1.5% in Precision over the original YOLOv8n, with only a slight increase in computational parameters (0.182M). The model's Recall rate decreases minimally by 0.01%. Compared to other models like SSD, Faster R-CNN, YOLOv5n, YOLOv7-tiny, YOLOv8n, and YOLOv10n, YOLOv8-Scm outperforms in mAP50, Precision, and Recall, and is significantly more efficient in terms of computational parameters. Specifically, the model achieves a mAP50 of 92.7%, a Precision of 86.6%, and a Recall of 87.2%. These results validate the model's superior capability in accurately detecting citrus sunburn across diverse and challenging natural scenarios. YOLOv8-Scm enables accurate, real-time citrus sunburn monitoring, providing strong technical support for smart orchard management and practical deployment.
BibTeX
@inproceedings{cong2025yolov,
  author = {Guo-xun Cong and Xinghong Chen and Zongyu Bing and Wenhuan Liu and Xiangling Chen and Yue Wang and Zheng Guo and Youliang Zheng},
  title = {{YOLOv8-Scm: an improved model for citrus fruit sunburn identification and classification in complex natural scenes}},
  booktitle = {Frontiers in Plant Science},
  year = {2025},
  doi = {10.3389/fpls.2025.1591989},
}
Figure