Detection of Apple Plant Diseases Using Leaf Images Through Convolutional Neural Network

Aim:

            To address the challenges associated with detecting apple plant leaf diseases using deep learning based models.

Synopsis:

            Plant diseases pose a significant threat to the global agriculture, leading to substantial crop losses. Detecting these diseases is challenging, often hindered by a lack of expert knowledge.  This work introduces a solution using deep learning, specifically a convolutional neural network (CNN), to identify diseases in apple leaves. The CNN model is designed with a smaller number of layers to address computational complexity. Augmentation techniques such as shift, shear, scaling, zoom and flipping are employed to expand the training set without acquiring more images. The model is trained on a public available dataset (Plant Village) for apple crops, focusing on identifying Scab, Black rot and Cedar rust diseases. Rigorous experiments demonstrate the efficacy of the proposed model achieving a high classification accuracy. Importantly, the model requires less storage and exhibits shorter execution times compared to several existing deep CNN models. The significance of the work lies in its suitability for the deployment on handheld devices, offering a practical and resource-efficient solution for real world application in agriculture. Despite comparable accuracy to other CNN models for crop disease detection, the proposed model stands out for its efficiency in storage and computational resource utilization.

Proposed System:

            Plant disease pose a significant threat to global agriculture, leading to substantial crop losses. The identification of these disease is a challenging task, often hindered by lack of expert knowledge. Deep learning based models, particularly convolutional neural networks (CNNs) offer promising solutions for disease detection using plant leaf images. However, the challenges such as the requirement for extensive training sets, computational complexity and the risk of overfitting persist in these techniques. In this study, a CNN is developed with a reduced number of layers to alleviate computational burdens. Augmentation techniques, including shift, shear, scaling, zoom and flipping are applied to augment the training set without the need for additional images. The CNN model is specifically trained for apple crops using the Plant Village dataset to identify Scab, Black rot and Cedar rust diseases in apple leaves. Experimental results demonstrate that the proposed model effectively identifies apple leaf diseases with a high classification accuracy. Notably, the model demands less storage and exhibits shorter execution times compared to several, existing deep CNN models. This makes it well suited for deployment in handheld devices, offering a practical and efficient solution for plant disease detection.