LT-YOLOv10n: A lightweight IoT-integrated deep learning model for real-time tomato leaf disease detection and management

Abstract

The challenges posed by a growing global population and the impacts of climate change have intensified concerns about food security, particularly in agriculture. This study introduces LT-YOLOv10n, a lightweight and efficient deep learning model tailored for detecting and localizing tomato leaf diseases. The model incorporates advanced architectural features, such as Convolutional Block Attention Modules (CBAM) and C3f layers, to improve feature extraction and emphasize disease-relevant areas. By adopting a reduced depth and width scaling approach, LT-YOLOv10n achieves high detection accuracy while ensuring computational efficiency, making it ideal for use on devices with limited resources.

The LT-YOLOv10n model was thoroughly tested against leading models, showcasing competitive performance in key areas like accuracy, inference speed, and parameter efficiency. It achieved an mAP50 of 98.7% on the test dataset and an inference speed of 87.28 FPS on the Jetson Orin Nano, demonstrating its effectiveness for real-time applications. Furthermore, the model was deployed in a mobile application, enabling real-time disease detection and offering actionable pesticide recommendations. The application integrates with ThingsBoard, an open-source IoT platform, to centralize prediction data and support remote monitoring through an intuitive dashboard. Features such as disease mapping, image-based predictions, and customized pesticide suggestions provide a comprehensive farm management solution.

This integrated system equips farmers with accessible, cost-effective tools for timely disease management, enhancing crop health and productivity while supporting sustainable agricultural practices. LT-YOLOv10n represents a significant step forward in agricultural AI, offering a scalable and efficient solution for precision farming.