A Deep Learning Approach using external feature fusion in fully connected layer for Plant Disease Recognition

Abstract

Plant Disease Recognition is very important for the food security and improvement in the production of sufficient food for the world by yielding proper treatment after diagnosing actual disease of the plant on time. This paper proposed a layered architecture of convolutional neural network to recognize the plant disease.  A deep learning architecture Alexnet has been used for disease recognition in which an external feature of segmented plant leaves is fused in the deepest fully connected layer of the deep convolutional neural network. This fusion of external feature of segmented plant disease to the Alexnet create a robust plant disease recognition system. Experimental analysis has been performed on a standard dataset PlantVillage which consists of 54,306 images of 38 plant diseases of 15 different plants. Proposed novel deep CNN performed well and outperformed to the state-of-the-art approaches.