An Improved YOLOv8 Model for Detecting Four Stages of Tomato Ripening and Its Application Deployment in a Greenhouse Environment

Haoran Sun, Qi Zheng, Weixiang Yao (), Junyong Wang, Changliang Liu, Huiduo Yu and Chunling Chen ()
Additional contact information
Haoran Sun: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China
Qi Zheng: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China
Weixiang Yao: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China
Junyong Wang: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China
Changliang Liu: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China
Huiduo Yu: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China
Chunling Chen: College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110866, China

Agriculture, 2025, vol. 15, issue 9, 1-33

Abstract: The ripeness of tomatoes is a critical factor influencing both their quality and yield. Currently, the accurate and efficient detection of tomato ripeness in greenhouse environments, along with the implementation of selective harvesting, has become a topic of significant research interest. In response to the current challenges, including the unclear segmentation of tomato ripeness stages, low recognition accuracy, and the limited deployment of mobile applications, this study provided a detailed classification of tomato ripeness stages. Through image processing techniques, the issue of class imbalance was addressed. Based on this, a model named GCSS-YOLO was proposed. Feature extraction was refined by introducing the RepNCSPELAN module, which is a lightweight alternative that reduces model size. A multi-dimensional feature neck network was integrated to enhance feature fusion, and three Semantic Feature Learning modules (SGE) were added before the detection head to minimize environmental interference. Further, Shape_IoU replaced CIoU as the loss function, prioritizing bounding box shape and size for improved detection accuracy. Experiments demonstrated GCSS-YOLO’s superiority, achieving an average mean average precision mAP50 of 85.3% and F1 score of 82.4%, outperforming the SSD, RT-DETR, and YOLO variants and advanced models like YOLO-TGI and SAG-YOLO. For practical deployment, this study deployed a mobile application developed using the NCNN framework on the Android platform. Upon evaluation, the model achieved an RMSE of 0.9045, an MAE of 0.4545, and an R 2 value of 0.9426, indicating strong performance.

Keywords: tomato detection; ripeness; YOLOv8s; android deployment; greenhouse environment (search for similar items in EconPapers)
JEL-codes: Q1 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17 Q18 (search for similar items in EconPapers)
Date: 2025
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2077-0472/15/9/936/pdf (application/pdf)
https://www.mdpi.com/2077-0472/15/9/936/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jagris:v:15:y:2025:i:9:p:936-:d:1642665

Access Statistics for this article

Agriculture is currently edited by Ms. Leda Xuan

More articles in Agriculture from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().