University of Florida researchers are building artificial intelligence tools that could help growers estimate strawberry and tomato yields more quickly and with greater precision.

The effort centers on two web-based applications, PhenoSeg and PhenoSnap, which are designed to process drone imagery and count crop features such as fruit and flowers. UF officials say the tools could give growers a faster alternative to manual scouting or reliance on historical records, both of which can be slow or imprecise when forecasting seasonal output.

The need for better forecasting is significant in Florida, where strawberries and tomatoes are major crops. According to the U.S. Department of Agriculture’s National Agricultural Statistics Service, the production value of Florida strawberries reached $714 million in 2025, while tomatoes were valued at $532 million.

Kevin Wang, an assistant professor of agricultural and biological engineering with the UF Institute of Food and Agricultural Sciences, outlined the project at an AgriTech conference in Plant City on May 5. Wang also authored a new Ask IFAS publication describing the tools.

PhenoSeg is the first step in the process. It uses drone images to separate individual strawberry plants from the surrounding background, allowing for plant-level analysis. PhenoSnap then counts fruit, flowers and runners on strawberry plants. It can also detect and count fruit and flowers on tomatoes.

The applications run on UF’s HiPerGator supercomputer, which researchers describe as the nation’s fastest university-owned system. Because the software is web-based, users do not need to install programs or own specialized computing equipment. They can upload images through a browser and receive results online.

During the 2025-26 growing season, researchers gathered drone imagery at the UF/IFAS Gulf Coast Research and Education Center and on two commercial farms. Wang said the early results are encouraging, particularly for plant segmentation in PhenoSeg. He also said the fruit and flower counting model in PhenoSnap is still undercounting in some cases, and that researchers are working to improve the algorithm in the next stage of development.

Drone imagery is a key part of the system. A drone flies over the field and captures high-resolution color images, covering more ground in less time than workers walking the rows or using a ground-based scouting platform. After the flight, the images are uploaded to PhenoSeg for plant separation, then to PhenoSnap for counting.

Wang is collaborating on the project with Wael Elwakil, a fruit and vegetable extension faculty member with UF/IFAS Extension in Hillsborough County, and Shinsuke Agehara, an associate professor of horticultural sciences at GCREC. He also credited UF strawberry breeder Vance Whitaker and UF tomato breeder Jessica Chitwood-Brown for helping shape the detection models used in the system.

The researchers want to expand the tools beyond the initial test sites and encourage commercial growers to try them. Wang said the team is open to working with growers to adapt the workflow for larger-scale use and to incorporate feedback that could improve the software.

Growers interested in learning more can contact Wang at [email protected].