The goal of this paper is to employ a fused convolutional neural network architecture to increase the accuracy of plant leaf disease identification.
Study Design: In this paper, we present a hybrid CNN architecture that adds a bio-inspired layer to an existing CNN design to increase accuracy and minimise the time required for leaf disease classification.
Between June 2018 to September 2020, at the National Institute of Electronics and Information Technology in Aurangabad.
Convolutional neural networks (CNNs) have become the de-facto approach for multi-dimensional data categorization. When it comes to this, activation functions like the rectified linear unit (ReLU), softmax, sigmoid, and others have shown to be quite useful. Additionally, standard CNN designs like as AlexNet, VGGNet, Google Net, and others aid this process by supplying standard and very effective network layer configurations. However, the enormous amount of calculations required to train and test the network limits the performance of these systems. Furthermore, as the number of classes grows, the network's validation and testing accuracy decreases. To address these issues, a hybrid CNN design was developed, which adds a bio-inspired layer to the original CNN architecture to increase leaf classification accuracy and speed.
Results: The suggested system was tested on a variety of leaf diseases, and it was discovered that it achieves greater than 98 percent accuracy in both testing and validation sets.
Conclusion: The most successful classifiers minimise the latency while increasing the accuracy. This motivates us to employ the suggested approach for real-time disease diagnosis in leaf images.
Please click here : https://journalcjast.com/index.php/CJAST/article/view/31099
No comments:
Post a Comment