An Efficient Approach for Crop Disease Detection using Deep Learning

Recognizing plant diseases is crucial for agricultural sustainability and high-quality produce. The conventional manual method of monitoring plant diseases is labor-intensive, requires expertise, and involves lengthy processing times. Despite advancements in machine learning (ML) and predictive analysis, the efficiency of classifying diseased plants remains limited, causing delays in report generation due to dataset constraints. Many researchers have developed crop disease detection models and soil testing models using various deep learning and machine learning approaches but encountered drawbacks such as poor robustness, high training time, and inaccurate accuracy. This research employs a deep-learning classification algorithm to distinguish healthy crops from diseased ones, aiming to provide a comprehensive report on the percentage of diseased and healthy crops, the specific type of disease, and corresponding remedies. The YOLOv5 model is chosen for its superior performance, achieving an accuracy of 0.90, surpassing manual sorting in identifying crop damage. The developed system includes a sensor-based device measuring soil pH, moisture, temperature, and TDS, ensuring optimal conditions for plant growth. Implementation of this method holds the potential to significantly increase crop yield and reduce losses. Notably, the YOLOv5 model stands out among other models due to its speed, accuracy, and proficiency in real-time applications. This model excels in detecting and classifying multiple objects within an image, making it particularly suitable for our plant disease classification task.