Classification of Rice-Plant Images into Healthy/Disease Class using ResNetV2 Variants
Swaetha Ramadasan, Manigandan Ramadasan, K. Vijayakumar, Gangadharam Balaji
Abstract
Deep-learning (DL) approaches have gained widespread adoption in diverse domains due to their enhanced accuracy and adaptability. Notably, the application of the pre-trained DL models is widely considered by researchers to solve a variety of image examination tasks. The proposed research aims to implement the pre-trained ResNetV2 (RN2) model for a chosen plant disease examination task. This work used RN2 to classify the rice-plant leaf images (RLI) into healthy and disease classes using the SoftMax classifier. This work implements various pre-trained RN2 schemes for organizing the photos using (i) conventional deep features and (ii) reduced features after a 50% dropout rate. The experimental outcome of this work confirms that the presented results demonstrate that the implemented technique provides better detection accuracy on the chosen images for both the conventional and the reduced features. This result verified that the implemented technique best examines the RLI.