Automatic Detection of Tomato Diseases using Fine-tuned pre-trained Deep Learning Models

Tomato is one of the most important crops in India. It has a high commercial value and is the second most widely produced crop. Diseases are harmful to crop health and have an impact on plant growth, either directly or indirectly. Plant growth must be monitored to ensure the minimum losses in production. There are numerous types of tomato diseases that deteriorate the quality of the tomatoes. As a result, the early crop treatment is crucial before it affects the entire crop. This paper presents a pre-trained convolutional neural network (CNN) based method for identifying and classifying the leaf diseases in tomato crops using transfer learning. The experiments are carried out on the Plant Village dataset, which includes ten tomato classes: Tomato Bacterial Spot, Early Blight, Late Blight, Leaf Mold, Septoria Leaf Spot, Spider Mites, Target Spot, Mosaic Virus, Yellow Leaf Curl Virus, and Healthy. Various pre-trained CNN models are fine-tuned with transfer learning approach. The tested pre-trained CNN models include DenseNet169, InceptionResNetV2, InceptionV3, VGG-16, VGG-19, DenseNet201, MobileNet, Mo-bileNetV2, and Xception. According to the results, MobileNet outperformed all other models with overall classification accuracy of 96%.