Berrybox: An affordable computer vision system for postharvest phenotyping of cranberry and other small fruits

Authors: Neyhart, Jeffrey; WAKHOLI, COLLINS - Oak Ridge Institute For Science And Education (ORISE); Rippner, Devin; Price, John; Altendorf, Kayla; Heineck, Garett; Zalapa, Juan; Loarca, Jenyne; SIDELI, GINA - Rutgers University

Submitted to: The Plant Phenome Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/27/2026
Publication Date: 7/1/2026
Citation: Neyhart, J.L., Wakholi, C., Rippner, D.A., Price, J.H., Altendorf, K.R., Heineck, G.C., Zalapa, J.E., Loarca, J.S., Sideli, G. 2026. Berrybox: An affordable computer vision system for postharvest phenotyping of cranberry and other small fruits. The Plant Phenome Journal. https://doi.org/10.1002/ppj2.70090.
DOI: https://doi.org/10.1002/ppj2.70090

Interpretive Summary: For cranberry growers and processors, berry color, shape, size, and uniformity are critical for ensuring the highest quality fruit. Additionally, quality is impacted by cranberry fruit rot, a disease that can cause yield loss of 100%. Selecting varieties from breeding programs with improvements in these attributes is possible, but current measurement methods are tedious and laborious. We developed a new, innovative, and low-cost system called the BerryBox for using AI models and pictures of cranberry to measure color, shape, size, and fruit rot incidence. This system uses readily available materials and drastically reduces the time and labor needed to measure berry characteristics. In tests of the BerryBox, we were able to identify and measure the shape and size of berries with over 99% accuracy, and the detection of fruit rot was over 98% accurate. We also showed the potentially broad applicability of the BerryBox by testing it using a sample of blueberry fruit with highly accurate results. This affordable and novel technology will greatly increase the efficiency of berry quality measurements in breeding programs of cranberry and other small fruits and will help lead to new varieties with ideal, uniform fruit characteristics.

Technical Abstract: Fruit size, shape, and color are important quality traits for breeding cranberry and other small fruits. The analysis of color (RGB) images can be used to rapidly measure these traits, but affordable hardware for standardized image capture and integrated user-friendly software pipelines are lacking. Additionally, no image analysis methods exist to estimate percent fruit rot, the most important disease impacting cranberry fruit quality and an otherwise tediously measured trait. We created the BerryBox, a simple and inexpensive lightbox and camera mount to standardize image capture of postharvest fruit samples. An accompanying publicly available software pipeline triggers image capture and color-correction, QR code processing, and output of data on berry size, shape, color, and percent fruit rot. Trained deep neural network models were highly accurate for segmenting sound fruit (F1 score: 99.4%) and detecting rotten fruit (F1 score: 98.5%). We applied the BerryBox to images of cranberries harvested from a breeding population of 156 half-sib clones across three years. Narrow-sense heritability ranged from 0.37 to 0.95, indicating that breeders can accurately select for favorable shape, color, size, disease resistance, and uniformity. Random subsampling of fruit showed that 25-30 berries were sufficient to describe the variation in the full dataset. We demonstrated the application of phenotypic data from the BerryBox in a small-scale linkage mapping analysis and detected significant marker-trait associations, showing that BerryBox traits capture underlying genetic variation. The software pipeline and BerryBox materials and assembly instructions are publicly available and are extensible to other similarly shaped fruits.