Design of a Predictive Measure to Enhance Neural Network Architecture for Plant Disease Detection

Agriculture is the foundation of Indian economic framework and financial exercises. It adds to national incomes, international trades, and overall industrial development. Indian population significantly depends upon agriculture for their livelihood. The key responsible reason in the decay of crop production is infected crops, which leads to noteworthy financial losses consistently. It is in this way absolutely critical to pursue the detection of leaf diseases. As per reports, the human populace is anticipated to reach at 9 billion by 2050, and food consumption is expected to increase by 60 percent. Improving and expanding the crop yields is hence a significant area of interest. Irresistible abiotic and biotic infections have essentially influenced potential yields and decreased it by a normal of 40%, a large number of population engaged in farming in this developing world suffering losses of up to 100%. Farmers around the globe are dealing with the diagnosis of plant leaf diseases and their legitimate treatments. The basic approach for classification process flows from data pre-processing to creating classifiers. Data pre-processing plays an important role in achieving higher classification accuracy while reducing performance time. In this paper, we analyzed the predictive measures of various artificial neural networks used in detection and classification of plant diseases, to better understand the existing architectures, and also proposed Gaussian filter as one of the predictive measure to increase the efficiency of trained model at optimal computational costs, while giving some insights about the future research work as well.