Exploring Deep Learning Approaches for Citrus Diseases Detection and Classification: A Review

Abstract

Citrus diseases pose significant threats to global agriculture, impacting crop yield and quality. In recent years the integration of deep learning models has surfaced as a hopeful method for classifying and detecting diseases. This review critically analyzes and synthesizes 25 research works that explore various deep learning models applications in citrus disease detection and classification. The methodology involves a systematic literature search, filtering based on relevance, publication date, and language. The selected works are categorized, and each is analyzed for contributions and limitations. The review identifies limitations, notably the reliance on limited datasets leading to issues of generalization and class imbalance. Data augmentation, while employed, lacks comprehensive evaluation. Practical implementation in real-world agricultural settings remains a challenge, demanding scalable, adaptable, and robust solutions. Future research directions are proposed to address limitations. Emphasis is placed on curating larger and diverse datasets, actively mitigating class imbalance, and rigorously evaluating data augmentation techniques.