<![CDATA[Tomatoes (Solanum lycopersicum) are one of Indonesia's mainstay horticultural commodities that are exported throughout Southeast Asia. However, the export value of tomatoes in 2021 recorded a decrease of 34.07% from 2020. The decline in the quality and quantity of tomatoes is generally caused by bacteria, fungi, viruses, and mite outbreaks that mostly attack the leaves such as late blight and two-spotted spider mite. This research utilizes one of the image processing methods to classify tomato leaf diseases in 3 labels, namely tomato healthy, tomato late blight, and tomato two-spotted spider mite. The image processing algorithm used in this research is Convolutional Neural Network (CNN) which can extract leaf image features in depth through its layer architecture. The VGG16 transfer learning architecture is used in this study because of its simple structure and can be modified by adding a fully connected layer, namely dropout with a value of 0.5 to adjust the model and improve its performance. Green Channel + CLAHE is also applied at the preprocessing stage with an epoch parameter of 30. The dataset used consists of 1,591 images of healthy tomato leaves, 1,909 images of late blight tomato leaves, and 1,676 images of two-spotted spider mite leaves. Two scenarios were conducted on the model, namely the model with callback function and the model without callback function. Based on the training and evaluation of the model that has been carried out, the model with the callback function is able to produce an accuracy value of 99.03% with precision for the labels tomato healthy 0.99, tomato late blight 1.00, and tomato two-spotted spider mite 0.98, and the number of incorrectly predicted images is only 15. This shows a higher value than the model without the callback function. Against 21 test images from other datasets, the model with callback function was able to produce accurate classification with high prediction values.]]>
