Plant breeding is a lengthy and demanding research. Traditional strawberry breeding requires many man-hours to manually measure plant characteristics, record data, and evaluate various desired traits. Also, human biases and prior perceptions or expectations can play a role in skewing the results. Thus, the plant breeding program at the University of Florida has developed AI tools to assist in different stages of breeding research. Developed AI models have offered accurate and quick data analysis to identify and quantify plant phenome (anatomical characteristics and traits). This reduction in the number of manhours to manually measure, record data, and perform destructive sampling, has greatly increased the ability to screen more breeding lines with fewer resources (time, plants, and money). These AI models can accurately with a high level of consistency measure the size of plant canopy, flowers, runners, and fruit maturity repeatedly throughout the season to create an individual profile of each tested breeding line. Five YOLOv8 based (computer vision) models were trained for strawberry runner detection including GI, UL-AI, SL-AI, Hybrid I (GI SL-AI), and Hybrid II (GI SL-AI UL-AI). Hybrid II model achieved 91% precision accuracy and 83% mAP50 (mean average precision at IoU of 50%). The use of AI image and video analysis has been reducing the time and resources needed to develop new varieties.
