Significance of Data Augmentation in Identifying Plant Diseases using Deep Learning
Thappidi Bhargavi, D Sumathi
Abstract
In recent years, rice infections have received an increasing amount of attention. Damage to rice plants leads to lower rice production. Identification of rice diseases early is essential for crops. However, conventional disease detection techniques are not very effective in doing so. Recent advances in convolutional neural networks have resulted in noticeably better images. Because of its excellent classification accuracy, it is particularly well suited for identifying various plant diseases. Plant diseases must be recognized when monitoring crops using technology-based methods. Recent research has demonstrated that CNN (Convolutional Neural Network) is the most efficient deep learning technique for processing leaf image data for illness diagnosis. Processing full leaves also increases computational cost and time, lowering training quality and performance. In order to increase the training sample size and boost classification accuracy, data augmentation is crucial. The classification of the enriched rice-leaf picture data set is thuscarried out in our work using the Resnet model. This method iseffective, as shown by the trained model’s accuracy of 98.10% on the test dataset. This research work has attempted to enhance theperformance of the model with and without data augmentation.