1. Evaluation of the convolutional neural network and the transfer learning algorithm for the purpose of improving the accuracy of rice leaf disease classification.
- Author
-
Yashwanth, K., Akilandeswari, A., and Sathish, K. S.
- Subjects
- *
CONVOLUTIONAL neural networks , *MACHINE learning , *RICE diseases & pests , *PLANT diseases , *AGRICULTURE - Abstract
The research aims to compare the Convolutional Neural Network (CNN) classifier method with Transfer Learning (TL) to detect different types of plant diseases from diverse agricultural regions. This component handles the gathering and processing of the dataset. A total of two study groups, CNN and TL are used. This research examined the disease in rice leaves using a publicly available data set with a sample size of 3,269 Rice leaf pictures (JPEG/JPG). The data was downloaded from the UCI Repository. The samples were separated into two groups, training samples (80 %) and test samples (20%). For the sample size calculation, the values are Alpha=0.05, Beta=0.02, and G power=80 %. Convolutional neural networks have a 94.88 % accuracy rate, whereas transfer learning has a 91.80 % accuracy rate. P has a significance value of 0.004 (p<0.05), indicating statistical importance. The rate at which the proposed algorithm identifies diseases is assessed and contrasted with the conventional method. The results of the accuracy comparison indicate that the accuracy of the Convolution Classifier (CNN) algorithm surpasses that of TL. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF