Researchers from University of Pennsylvania state in cooperation with the staff of the Swiss Federal Technological Institute built a model of the system, connecting multiple computers into the neural network. Then they began to train the system using a database consisting of more than 53000 photos, which were sealed with both sick and healthy plants. It is worth noting that the database included 14 types of main agriculture crops and 26 the most common diseases.

Using machine learning technology, the researchers trained the system to find correspondences between plants and diseases. As a result, developers have managed to teach it to identify plants diseases from photographs with an accuracy of 99.35 percent.