Deep Learning for Plant Leaf Disease Detection: A Comprehensive Review of Techniques and Early Detection Applications

Abstract

Abstract: Due to the growing global population and the ensuing rise in food consumption, agriculture continues to be every nation's top priority and most recent research topic. Particularly in applications involving visual data, audio data, and text analysis, developments in autonomous learning and feature extraction have drawn both academics and industry have shown a lot of interest. Agriculture plant protection, more specifically plant disease detection, has drawn a lot of interest among these applications. The subjectivity involved in disease spot feature selection can be addressed by using methods like Deep Learning (DL) and Machine Learning (ML) methods, which will result in more objective feature extraction and quick technological adoption. This review includes the works published between the year 2014 to 2023 that were acquired from dependable databases like Scopus and Web of Science. One hundred and three peer-reviewed articles were examined using search terms like "plant leaf disease diagnosis", "ML and DL methods." In this review, categorization techniques for plant diseases are systematically compared, with a focus on ML algorithms exactly Convolutional Neural Network (CNN) algorithm. The various aspects, such as experimental setup, metrics are considered for disease classification, the processing methods for each algorithm including image segmentation and feature extraction are considered in this study. Researchers looking to recognize certain plant diseases using data-driven methodologies will find the systematic comparison of techniques to be very helpful in highlighting the advantages and disadvantages of various approaches.  Furthermore, this study points towards promising future research directions, including the exploration of novel ML and DL architectures, integration of multi-modal data for improved accuracy, and the development of robust models to tackle complex disease scenarios. By leveraging the knowledge presented in this study, researchers can further enhance the detection of disease methods and back to the advancement area of agriculture.