Worldwide food production is being stressed by extreme weather conditions, fluctuating temperatures, and global affairs. With a global output of millions of tons annually, tomatoes stand as a pivotal staple in agricultural practices worldwide. Early-stage identification and classification of diseases in tomato plants can be a cost-effective measure for farmers, potentially reducing the need for expensive crop sprays and enhancing overall food yield. In the realm of disease detection and control, there is considerable potential for transformative impact through technological innovations. In a multitude of domains, deep learning algorithms, a subset of artificial intelligence, have autonomously demonstrated their recognition and applicability in real-life situations. This paper seeks to employ deep transfer learning for the classification of various distinct tomato diseases namely, bacterial spot, early blight, late blight, leaf mold, mosaic virus, septoria leaf spot, target spot, and yellow leaf curl virus with the healthy state. The approach in this work uses tomato leaf images as input, which is given to convolutional neural network architectures. In addition, these models utilize transfer learning principles from well-established deep learning networks. The assessment of performance involved rigorous examination through multiple data split strategies and diverse metrics. Moreover, to mitigate the influence of randomness, the experiments were repeated 6 times. The six categories were classified with mean values of 98.3% precision, 98.2% F1 score, 98.1% recall, and 98.4% accuracy.

#