Tomato leaf disease detection using series of Convolutional and Depthwise Convolutional Layers

Numerous studies have focused on enhancing the effectiveness of identifying leaf diseases through image classification. However, it is essential to develop a classification system with fewer parameters to enable it to operate efficiently on mobile devices. As a result, A lot of research works are going on to make the neural network computationally light so that we can utilise these networks on a mobile device as it cannot afford a GPU to run in background because of the space and memory limitations of a portable device. In this study, we propose a deep learningbased approach for tomato leaf disease detection using a series of convolutional and depthwise convolutional layers. The proposed model contains only 17,209 trainable parameters. The model was able to achieve high accuracy of 92.10 % on tomato crop from a publicly available PlantVillage dataset while utilizing a smaller number of parameters.