Effective Skip-based Residual Network Algorithm for Detection and Classification of Areca Plant Diseases

Abstract

Areca plant is a primary commercial crop that plays a vital role in providing economic security to numerous people in India. However, weather factors, particularly temperature and rainfall, highly impact the severity of disease and spread of infection in these crops. Significant development has been obtained by utilizing Deep Learning (DL) based algorithms in recent times. The classification of different Areca plant diseases is challenging due to the noise present in their images, as well as the complex layers of the models. In this paper, an Effective Skip-based Residual Network (ESkip-ResNet) algorithm is developed to detect and classify Areca plant diseases. The ESkip-ResNet architecture includes skip connections with ResNet, which contain many phases and leftover blocks to ensure improved classification performance. The ESkip-ResNet utilizes rest blocks to identity mapping by skipping connections in the ResNet architecture. Moreover, ESkip-ResNet incorporates an efficient approach for downsampling and stabilizing batch normalization layers to enhance the model’s stability and classification performance. The developed ESkip-Resnet algorithm obtains 97.47% accuracy on the Arecanut dataset and 99.98% accuracy on the Plant Village dataset, outperforming conventional algorithms.