Proof of Concept: Creation of a Computer Vision Dataset for Tomato Brown Rugose Fruit Virus Detection

Tomato brown rugose fruit virus (ToBRFV) is a rapidly spreading RNA virus that severely impacts tomato and pepper crops worldwide, causing significant losses since its identification in 2014. Its rapid spread across open and protected fields underscores the need for early detection, as traditional visual inspections are time-intensive and prone to errors. This research uses vision algorithms and deep learning to enhance detection efficiency and address the shortage of skilled operators. A new, augmented ToBRFV dataset was developed, achieving significantly higher precision and recall than models trained on the original 242 images. Two practical prototypes were introduced for testing the dataset: a real-time detection system using the NVIDIA Jetson Orin for rapid, in-field applications and a web-based platform allowing users to upload and analyze images remotely. The Jetson Orin system demonstrated low-latency, high-speed inference, which is ideal for greenhouses and cultivation fields, while the web-based solution provided accessibility for users with limited hardware resources. These advancements offer scalable, efficient, and accurate tools for early ToBRFV detection. At this early research stage, the development aims to be expanded for seed detection and monitoring plant growth stages, facilitating precision agriculture and minimizing crop losses.