An Enhanced Deep Learning Algorithms for Image Recognition and Plant Leaf Disease Detection

Object recognition in photographs has only been possible with the advent of modern deep learning (DL) techniques. Following its success in other fields, DL techniques are increasingly being used to a broad variety of agricultural operations. Due to its intrinsic complexity, image-based plant disease detection is difficult to solve. Targeted management relies on more than just detection and requires accurate identification of each species. An automated disease detection system aids the farmer by providing a speedy and accurate diagnosis of plant illness. Accelerating agricultural diagnostics requires the development of an automated system for identifying plant leaf diseases. Using machine learning and image processing, the authors of this study outline a strategy for detecting leaf blight. A photo of a leaf may be fed into this framework. In a first processing stage, noise is eliminated from the leaf images. The mean filter may be used to get rid of noise. Histogram equalisation is a method for enhancing image quality. Segmentation is a technique used in photography to break a large image down into more manageable sections. This helps to establish the picture's limits. The image is broken up using the BIRCH (Balance Iterative Reducing and Clustering utilising Hierarchies) technique. For this purpose of feature extraction, fine grained classification model, a convolutional neural network is used. After that, the pictures are categorized using a variety of methods, one of which is RCNN (Region based Convolutional neural network). Using this approach, this study discovered the split ratio, computational time, and as a result RCNN has resulted in a high accuracy of 0.98 when detecting the leaf compared to other algorithms.