Advancements in Plant Leaf Disease Recognition: YOLO-Based Deep Learning Approaches
: Plant leaf diseases cause severe losses in crop yields and qualities, and account for considerable volume of losses to the agricultural output globally. Recognition of plant disease early and rightly is crucial to disease treatment and to reduce loss to the crop and to maintain agricultural sustainability. Plant disease that occurs on the leaves has been traditionally detected by farmers and experts with naked eyes by checking its symptoms like discoloration, spots and lesions. However, the process requires time, labour, expertise and is subjective, which renders it unusable for large-scale implemented agriculture. Recent years have seen the promising use of Artificial Intelligence (AI) as a tool for automated plant disease identification. The extraction of manually-crafted features from photographs of plant leaves, such as colour, texture, and form, is at the heart of many Machine Learning (ML) approaches used for disease classification. While these ML models have shown acceptable performance, they require significant manual feature engineering and can be poor at operating in real-world settings and with voluminous data. To address these issues, Deep Learning (DL) algorithms have found extensive usage in the identification and categorisation of plant leaf diseases. The You Only Look Once (YOLO) family of detection of objects models is making waves in the DL object detection space thanks to its impressive dual-tasking capabilities: object identification and multiple illness categorisation in a single pass, all at lightning speed and with pinpoint accuracy. For real-time disease identification in precision agriculture, YOLO stands out as an end-to-end feature learning and object recognition method, set apart from typical ML approaches. Understanding the DL models suggested for plant leaf disease detection and classification using the YOLO principle is the primary goal of this survey. It also provides a comparative and performance analysis of these models by examining their techniques, merits, demerits, datasets used, and evaluation metrics.