Manuscript Title:

FINE TUNING A DEEP CONVOLUTIONAL NEURAL NETWORK WITH NATURE INSPIRED CAT SWARM OPTIMIZATION IN ORDER TO CATEGORIZE ORYZA SATIVA DISEA

Author:

NV RAJAREDDY GOLUGURI, SUGANYA DEVI K, HEMANTH KUMAR V

DOI Number:

DOI:10.17605/OSF.IO/KHX2W

Published : 2021-12-17

About the author(s)

1. NV RAJAREDDY GOLUGURI - Computer Science and Engineering, National Institute of Technology Silchar, Ghungoor, Silchar, Assam -788001, India.
2. SUGANYA DEVI K - Computer Science and Engineering, National Institute of Technology Silchar, Ghungoor, Silchar, Assam -788001, India.
3. HEMANTH KUMAR V - Computer Science and Technology, Raghu Institute of Technology, Dakamarri, Visakhapatnam,Andhra Pradesh-531162, India.

Full Text : PDF

Abstract

Bacterial Leaf Blight (BLB), Brown Spot (BS), and Leaf Blast(LB) are all diseases that can damage Oryza Sativa. Plant diseases of the Oryza Sativa species are causing farmers to make a loss. This illness must be diagnosed in order to avoid more significant economic losses. To accomplish the first detection, each plant was visually inspected, which proved to be a difficult process. Our goal was to create a new and more successful approach for detecting and diagnosing BLB, BS, and LB in Oryza Sativa plants. In this work, image denoising is accomplished using the wavelet soft thresholding technique.In this example, the Salp Swarm Algorithm (SSA) is employed to determine the optimum threshold value to maximise the Peak Signal to Noise Ratio (PSNR). Following that, the images are segmented using the Enhanced K-Means (EKM) clustering technique. Finally, the Deep Convolutional Neural Network is used to detect and classify diseased Oryza Sativa images (DCNN). To choose the best weights for DCNN, the Cat Swarm Optimization (CSO) method is employed.The findings of this study reveal that the selected technique obtains the highest accuracy, 96.9%, when compared to the current DCNN-Artificial Fish Swarm Optimization(AFSO), DCNN-Artificial Bee Colony(ABC), and DCNN-Particle Swarm Optimization(PSO) approaches.


Keywords

Deep Convolutional Neural Network, Rice Diseases, Cat Swarm Optimization, Enhanced K-means Clustering, Salp Swarm Algorithm.