Real-Time Tomato Leaf Disease Classification for Vertical Farming: Training on Public Data and Preliminary Field Validation
Year: 2026
Type: conference-paper
Abstract: Vertical and hydroponic farming techniques necessitate computerized monitoring of plant health for the early diagnosis of diseases. Collecting disease training data from operational indoor farms is difficult due to the regulated settings that generally reduce disease prevalence. This study introduces a real-time classification system for tomato leaf diseases utilizing the YOLO11s deep learning model, intended for implementation in vertical farming environments lacking farm-specific disease data. The model was trained on the Tomato Leaf Disease dataset, which includes 11 classes (10 diseases and one healthy category), achieving a Top 1 accuracy of 98.61 % and a macro$F 1$-score of 98.70 % on the benchmark test set, surpassing ResNet-50, EfficientNet-B0, and YOLOv8 under the same conditions. To assess deployment viability, we evaluated the model using photos obtained from an operational vertical hydroponic farm. The model attained 97.8 % accuracy in recognizing healthy leaves in real indoor circumstances, indicating effective generalization across the domain gap between laboratory training data and agricultural application. The solution facilitates real-time monitoring for automated farming pipelines with an inference latency of 5.89 ms per image.
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W7168259101