A Critical Analysis of Convolutional Neural Networks for Leaf Disease Detection in Plants

Deep learning techniques have just recently been used to picture identification. Convolutional neural networks, a sort of deep learning technology, have shown outstanding breakthroughs in this discipline. Convolutional neural networks have shown effective in applications such as object identification, face recognition, handwritten digit recognition, and traffic sign recognition. The success of CNN in image recognition has prompted researchers to investigate their application in agriculture for a broader range of purposes, including plant identification, crop monitoring, weed detection, water and soil management, fruit counting, disease diagnosis, and crop health assessment. The availability of different research works on the use of deep learning models in agriculture makes it difficult to pick an appropriate model based on the kind of dataset and experimental setting. This paper focuses on the plant disease prediction using leaf pictures collected with a deep convolutional neural network. This paper also examines and discusses several preprocessing approaches, convolutional neural network models, frameworks, and optimization strategies for recognising and categorising plant illnesses in their leaves. We also discusses the datasets and performance indicators that are used to assess the efficacy of a model. The strengths and downsides of various techniques and models discovered in the existing research are highlighted. This survey will be useful to researchers interested in applying deep learning approaches to the problem of leaf disease diagnoses.