A Fine-Tuned Convolutional Neural Network Model for Banana Leaf Disease Detection

Abstract

The use of Convolutional Neural Networks (CNNs) for early identification of banana leaf diseases is a significant advancement in the area of image recognition technology. The primary objective of the proposed work is to develop and implement a CNN model that is capable of reliably locating and recognising banana leaves in images. To implement this strategy, a huge dataset of labelled images of banana leaves are compiled. These images have been selected from Kaggle and preprocessed to ensure that they are both varied and regular. In the pre-processing step, the images have been transformed into identical sizes to be understood by the CNN model. The capabilities of CNNs to learn hierarchically are used to construct a multi-layered architecture that includes convolutional, pooling, and fully connected layers. As part of the training process for the model, the dataset is subjected to recurrent optimisation to reduce a predetermined loss function. The capability of the model to identify banana leaf diseases in a broad variety of images is put to the test and validated via a series of layers of the CNN model. It has been shown that CNN is capable of accurately locating banana leaf diseases by achieving an accuracy of 98.24% at epoch 20. Further, the loss of the model has been also identified as the lowest at epoch 20 as 0.2936. This ability has the potential to be useful in a variety of fields, including agriculture, cookery, and environmental awareness. The development and validation of this CNN model's ability to identify banana leaf disease is a significant advancement in the area of automated visual recognition systems. This model has the potential to be used in a wide variety of industries and organisational applications.

Topics & Concepts