The identification of diseases in plants is crucial for sustaining crop yield and securing global food resources. This study delves into the utilization of deep learning methodologies to precisely detect diseases in tomato plant leaves. Focused on the PlantVillage dataset, a comprehensive collection of tomato plant leaf images categorized as healthy or diseased is analyzed. Augmentation techniques, encompassing random transformations like rotations, flips, and zooms, are employed to diversify the training dataset, aiming to improve the models’ accuracy and generalization capacities. The deep learning models, including AlexNet, DenseNet, LeNet, and two bespoke models, undergo training with well-suited loss functions and are optimized using appropriate techniques. Assessment of the models’ performance involves metrics such as accuracy, loss, recall, precision, and F1 score, calculated on distinct validation sets. The outcomes indicate that the specifically tailored models for disease detection outperform established architectures. Notably, Custom Model 2 achieves an accuracy of 86.10% and an F1 score of 85%, while Custom Model 3 attains an accuracy of 86.84% and an F1 score of 86%. Contrastingly, AlexNet records an accuracy of 79% and an F1 score of 79.29%, DenseNet exhibits an accuracy of 89% and an F1 score of 88%, and LeNet scores an accuracy of 81% and an F1 score of 82%. This study underscores the potential of deep learning in precisely identifying plant diseases in tomato leaves, contributing to the advancement of sophisticated diagnostic techniques. Implementing these models in agricultural systems holds promise for proactive disease identification, safeguarding crops, and ensuring global food stability.