An AI tool has been developed to identify olive varieties from images of olive pits, known as ‘OliVaR’. The neural network was trained using an extensive database of olive fruit endocarps (olive pits), compiled by partners of the GEN4OLIVE European project. With contributions from five germplasm banks across different countries and advancements in AI technology, particularly from the University of Cordoba, which provided significant data, this tool promises to revolutionise the identification process.

Under the coordination of the Ucolivo group from the María de Maeztu Unit of Excellence – Department of Agronomy (DAUCO), olive germplasm banks from various Mediterranean countries collaborated to collect over 150,000 photos of 133 olive varieties. The Computer Science Department at Sapienza University in Rome spearheaded the data collection and algorithm creation, introducing a novel method to automate traditional morphological classification.

Researchers Hristofor Miho and Concepción Muñoz Díez emphasised the model’s accuracy, boasting around 90% efficiency. They highlighted the machine’s learning approach, which continuously improves through trial and error. By adhering to strict protocols and unifying methodologies, project participants optimised the algorithm’s performance.

The outcome is an AI tool capable of detecting intricate morphological details imperceptible to the human eye. Upon analysis, it provides a list of potential olive varieties with varying degrees of compatibility with the input sample. This Machine Learning system will serve as the foundation for the user-friendly application, enabling growers and nurseries to swiftly identify olive tree varieties.

By offering this tool publicly and free of charge, Ucolivo aims to enhance the collective understanding of olive varieties within the industry.