Developing an app that can Identify Olive Tree Varieties from Stone Photos thanks to the largest photographic database of olive fruit endocarps. The development of this tool – generated by the European project for the improvement of olive oil GEN4OLIVE – was made possible thanks to the cataloguing and documentation work of five germplasm banks of different countries and to advances in artificial intelligence systems, as reported by theUniversity of Cordoba (UCO).
The initiative saw the participation of olive gene banks from Spain, Italy, Greece, Morocco e Turkey committed to raising more than 150.000 foto di 133 varieties of olive trees of the Mediterranean basin.
Il Department of Computer Science, Sapienza University of Rome was tasked with gathering information and create the algorithm for this tool, which proposes a new approach to identify varieties and automate the traditional morphological classification process.
Effectiveness equal to 90%
The researchers explained it Christoph Miho e Conception Munoz Diez, which highlighted the precision demonstrated by the model, with aeffectiveness of approximately 90%. "It deals with – they stated to Mercacei - of a system based on machine learning, made of trial and error, in which we train the machine to acquire ever greater precision by learning from its own mistakes”.
The researchers pointed out that the more images are part of the database, the more effective the system will be. All the participants in the project have agreed on very strict protocols to unify their working methodologies and generate images that allow the optimization of the algorithm.
The result is an Artificial Intelligence that has proven capable of detecting morphological details that even escape the human eye. After data processing, The app provides a list of possible varieties which present different degrees of compatibility with the photographed sample.
It will be made public and free
This system of machine learning algorithm will be the basis of an application that will allow farmers or nurserymen to easily and quickly identify the variety of olive tree they are working with. The researchers assured that by making it available to the entire industry as a free and public tool, this app will also contribute “to advance general knowledge of all existing olive varieties”.
