Matsuyama, Japan - August 05, 2024

[Credit: M.P. Islam, Ehime University]

Researchers have developed an innovative artificial intelligence (AI) system aimed at revolutionizing the monitoring of tomato plants in plant factories. The system, based on a universal multi-branch general-purpose convolutional neural network, promises advanced capabilities in disease detection and plant health management.

Real-time monitoring is crucial in plant factories to swiftly identify and classify diseases, thereby preventing potential outbreaks. The proposed DeepD381v4plus network has demonstrated exceptional performance, boasting high accuracy scores across various tomato leaf diseases, including sensitivity, specificity, precision, F1 score, and Matthews correlation coefficient, all surpassing 0.96.

In addition to disease surveillance, the AI system monitors critical stages of tomato plant growth, such as bud formation, flower appearance, and fruit set, crucial for confirming successful pollination. The detection module, DeepDet381v4 – YOLOv4M, achieves impressive results in identifying and counting ripe tomatoes from a distance of 40 cm with minimal error, enhancing efficiency in harvest management.

Both AI networks operate efficiently, with classification and detection tasks performed at over 27 frames per second (fps), indicating high-speed performance despite their compact size.

The experimental approach not only aids in disease prevention and accurate fruit detection but also promises to reduce labor costs, optimize cultivation practices, and ensure the harvesting of tomatoes at their peak maturity. This advancement is poised to significantly elevate crop management standards and enhance tomato yield quality for farmers.