Plant Quality Assessment and Disease Identification System Using AI

Abstract

Agriculture is the pillar of the economy and salient to accomplish comprehensive development. Presently, plant disease detection and quality assessment has encountered a rising scrutiny as a large dataset of plants is produced by monitoring its features. The naked eye inspection is the conservative approach endorsed in practice for plant leaves disease identification and inspection. Besides being subjective, it is becoming impracticable. This approach is followed by traditional machine learning methods for classification. In this paper, all the AI techniques from traditional to the most recent ones are concerned. Among all image classification techniques, we studied which method and algorithm surmounted every other and deployed it. Crop quality appraisal is an extremely fundamental assignment as it can assume an indispensable part in the understanding of the nature of the harvests and its sicknesses. Computer-based intelligence and ML is helping make that objective conceivable. Nowadays, deep learning and GAN-based techniques have been used for the quality appraisal of harvests and disease detection. Preliminary knowledge on plant health and disease identification will result in advancements in AI and management strategies. It has paved a way for constructing trained models with higher accuracy and effectiveness. This paper comprises various phases like image acquisition, dataset preparation, training, validation, and evaluation. Overall, all these play a pivotal role in shaping the agriculture infrastructure on a gigantic global scale. Iterating the training model by tuning different parameters has resulted in optimal training and validation accuracy and least training and validation loss. The final experimental results indicate that training accuracy achieved is approximately 98.5% followed by a training loss equivalent to zero and remains constant with respect to the epochs.