The most agriculturally dependent countries are striving to use the current technological revolution to revolutionize the agricultural sector and to monitor crops with the utmost precision in the shortest possible time. It is crucial to have a technique to automatically classify leaf diseases too quickly and accurately. A lot of research on plant diseases classification has been done using CNN, decision trees, or neural networking-based classifiers. The main objective of this research is to carry out a noble approach of tomato leaf disease classification based on EfficientNet-B0 which can accurately classify tomato leaf diseases with maximum accuracy. Initially, Features are extracted from images using basic EfficientNet-B0 architecture, then the features are sent through a set of fully connected, dropout and batch normalization layer which are also appended by Rectified Linear Unit activation function. Inclusion of extra layers and activation functions at the end has fine-tuned the model which is reflected on the output of the model. A tomato Leaf dataset with 25851 images from Kaggle was collected to test and train the proposed model. Also, around thousands of images has been collected from the field to test the model. Utilizing Kaggle as the compiler, it was possible to achieve an accuracy of 99% with the proposed method. This proposed approach performed much better compared to original EfficientNet-B0, MobileNet-V2, and VGG-19. This procedure will escalate much beyond the scope of the previous research.