Tomato brown rugose fruit virus (ToBRFV) in Top-View Images of Tomato Plants Using Deep Learning

Tomato (Solanum lycopersicum) is one of the main staple foods worldwide and is a strategically important agricultural product due to its economic and nutritional value. However, tomato production may be affected by various plant diseases. One of the most important disease is tomato brown rugose fruit virus (ToBRFV). Consequently, the early detection of ToBRFV infection is critically important for mitigating yield losses and supporting sustainable agriculture. This study aims to detect ToBRFV from top-view RGB camera images of commercial tomato plants grown in a greenhouse environment by using deep learning. As a part of a scientific research project, 40 healthy and 45 ToBRFV inoculated commercial tomato plants were grown in two controlled greenhouses. Top-view video recordings were captured over a 30-day period. The dataset was compiled by selecting the highest quality frames from these video recordings. In order to minimize the influence of environmental factors in the images, segmentation was applied and the background was converted to black. The compiled dataset was partitioned into 60% for training, 20% for validation, and 20% for testing. For early detection, ResNet-50 was chosen due to its deep architecture, residual connections, superior feature extraction capability, and robustness. A total of 4054 healthy and 16747 ToBRFV inoculated commercial tomato images were used for 5-fold cross-validation. The cross-validation results yielded an accuracy of 0.9601, precision of 0.9275, recall of 0.9617, and an F1-score of 0.9410. The accuracy variance was calculated as 6.56, while the precision variance was 24.67. Recall demonstrated a more consistent performance with a variance of 2.79. These results indicate that the model exhibits stable performance. The findings show that the ResNet-50 model achieves high classification performance and provides high accuracy in early detection of ToBRFV infection. This study makes important contributions towards fostering sustainable agriculture and reducing yield losses associated with this disease.