Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Highlights

An iOS-based app (BlueberryCounter) was developed for blueberry detection, counting, and maturity assessment.

Abstract. Harvest maturity and yield estimation of blueberries are important for growers to optimize crop production and stay competitive. It is extremely labor-intensive and infeasible to assess fruit maturity and count fruit for yield estimation manually. The pervasive use of smartphones and the recent advancements in deep learning and edge computing open new opportunities for automated, inexpensive approaches to image-based blueberry detection and counting. This study presents a new, simple mobile application (app) developed using Swift in iOS (version 16), i.e., BlueberryCounter, which enables growers to assess fruit maturity and count rapidly. The app features a user-friendly interface and supports real-time blueberry detection, counting, and maturity estimation based on two YOLOv8-based fruit detectors: YOLOv8m (Fast) and YOLOv8l (Accurate), which offer users flexibility in choosing between speed and accuracy. A live window visualizes real-time detection results alongside the detection logging statistics displayed. The fruit detectors deployed on the app were trained and evaluated on a blueberry canopy image dataset consisting of 17,809 annotated blueberry instances, including 6,958 instances of ripe (“Blue”) and 10,851 unripe (“Unblue”) fruit. The in-season testing showed better accuracy than the cross-season testing, implying more efforts are needed to improve model robustness across different seasons. The operational functionalities of the app were verified on a smartphone in preliminary field testing. The BlueberryCounter, which will be made publicly available, promises to evolve into a useful tool on mobile devices for blueberry growers.