A Comprehensive Review: Advancements in Pretrained and Deep Learning Methods in the Disease Detection of Rice Plants

Abstract

Rice cultivation is essential to the global economy, particularly in India, where it holds the distinction of being the largest rice exporter and the second-largest rice producer. However, the agricultural sector faces significant challenges due to diseases and pests that negatively impact the crops, by hindering the plant growth, reducing the yield, and, in extreme cases, leading to famine. The use of pesticides, intended to increase production, often results in a decline in crop quality. Prompt as well as precise disease identification in plants is requisite for prevention and control of disease, enabling timely implementation of pesticide control measures. This has spurred research at the intersection of computer science and agriculture, specifically focused on identifying diseases in rice through collected and real-time images. Deep learning (DL) has emerged as a key area of study within this domain, addressing various aspects of agricultural plant protection, including disease detection and pest control. Pretrained models have proven to be invaluable tools in the realm of rice plant disease identification and monitoring. These models leverage transfer learning, enhance feature extraction, reduce training time and resource requirements, improve generalization and resilience, and facilitate knowledge sharing and collaboration. This article examines rice plant diseases, explores deep learning and pre-trained models for diagnosis, reviews relevant publications, and presents a comparative analysis of research studies to assess advancements in rice plant disease detection.