Plant Leaf Disease Classification Using Fine-Tuned Pre-Trained CNN Models

Plant disease detection is crucial for ensuring agricultural productivity and preventing crop loss. This study explores shallow retraining of pre-trained deep learning models (ResNet-50, EfficientNet-B0, MobileNetV2, and AlexNet) for plant disease classification using plant leaves. We implemented our models using tomato leaves from the well-known PlantVillage dataset. To enhance model performance, we applied image preprocessing, dataset balancing, and batch size optimization, ensuring efficient learning. Results from experiments conducted on the PlantVillage dataset show that ResNet-50 achieved the highest accuracy (97.32%) but required the longest training time. Conversely, EfficientNet-B0 and MobileNetV2 demonstrated a balanced trade-off between accuracy and computational efficiency.