Nutrient Deficiency Prediction in Banana Leaves Using Advanced UNET Architecture

India is agriculture-based country, and the maximum population depends on agricultural productivity for their wealth. India is the second leading country in farming and agriculture is the backbone of economic development. The need for agricultural products is increasing day by day, and at the same time the changing environment produces so many changes in biodiversity, which influences the farming and productivity of plants. The growth and health of plants mainly depend on the soil nutrients and most of the time it’s not easily identified by farmers and experts. It is necessary to take suitable remedial actions such as adjusting fertilizer usability, and improving the soil quality to optimize plant growth and avoid nutrient deficiency in plant leaves. So, it needs support from computer-aided techniques that predict the nutrient deficiency plants of plant leaves easily. This study adapts Deep Convolutional Generative Adversarial Networks (DCGAN) for data augmentation, Denosing AutoEncoder (DAE) for noise removal, and used UNET, Advanced UNET for nutrient deficiency prediction. This proposed model experimented with the Mendeley Banana Nutrient deficiency dataset and achieved an accuracy of 99.18% for Advanced UNET, and 97.09% for UNET. This experiment recommends Advanced UNET for banana leaf nutrient deficiency detection.