Automatic Classification of Plant Leaf Images into Healthy and Disease Class with EfficientNet: A Study

To handle the variety of digital data, a significant number of automatic data examination techniques have been created recently. These algorithms are essential for analyzing image data in many different fields, including agriculture. One frequent activity in agriculture is plant health monitoring using image processing, and the goal of this research is to provide a way for more accurately classifying plant leaf data into the healthy and disease classes. Data on tomato plant leaves were selected for this investigation. This system consists of three stages: binary classification with 3-fold cross validation and verification, deep feature mining with a selected algorithm, and image collection and resizing. The pre-processed image helps to obtain an enhanced outcome compared to the raw leaf data, according to the experimental results of this work, which is conducted utilizing the selected pre-trained models employing the raw and pre-processed photos. In this study, a binary classification utilizing SoftMax is implemented. The detection accuracy of the data, both raw and pre-processed using adaptive thresholding, is >88% and >92%, respectively. This study validates that, when applied to the selected leaf data, the suggested technique yields superior results.