Greenhouse tomato fruit recognition based on CBAM-YOLOv8
LIAO Xinzhi
KONG Guoxi
LIN Guichao
CAO Huiru
LI Xiaomin
Abstract：Efficient and accurate detection of tomato fruits is the key to achieving intelligent harvesting.To address the challenge of balancing accuracy and speed in existing deep learning object dectection models,this study proposed an im-proved YOLOv8 model.This model incroporated the CBAM attention module during the YOLOv8 feature extraction stage,which dynamically adjusted feature weights to effectively suppress noise and irrelevant information,and improved the accuracy of model detection.The experiments showed that CBAM-YOLOv8 performed well in tomato detection,with accuracy,recall,and average precision reaching 91％,78％,and 90％,respectively.Compared with SSD,Faster RCNN,and original YOLOv8,the performance had significantly improved.This model effectively reduced the rates of false positives and false negatives.In terms of prediction time,YOLOv8 has the shortest comprehensive time consump-tion.In contrast,the prediction time required by CBAM-YOLOv8 model has increased,and the inference speed is slower,thereby increasing the computational cost.Therefore,in practical applications,a balance needs to be made between accu-racy and speed,In conclusion,CBAM-YOLOv8 provides an effective solution for real-time monitoring,yield estimation,and efficient harvesting of tomato fruits.
Keywords：Tomato recognitionTarget detectionAttention mechanismCBAM-YOLOv8
Publication Date：2025-09-05
Online Publishing Date：2025-09-29(First online date of this platform, not the publication date of the document)
Pages：9( 48-56 )