An Effective Rice Diseases Detection and Classification Using YOLOv8 Architecture with ChatCrop based Mobile Application

Rice is a vital staple crop worldwide, but it faces challenges from diseases like Brown Spot, Leaf Blast, Bacterial Leaf Blight (BLB), and Tungro, reducing both yield and quality. Conventionally, disease detection has been dependent on human visual inspection, which is time-consuming and error-prone. The ERDCYC model envisions a device that can do automatic detection in real-time, an initial dataset features 8,883 images of five leaf categories: healthy, brown spot, leaf blast, BLB, and tungro, with enhanced data quality through image preprocessing and augmentation techniques, while transfer learning on a pretrained ImageNet model is done to facilitate feature extraction. Disease detection is furthered in great detail and accuracy by the new model architecture of YOLOv8, which is chosen due to its speed and precision. With the help of Task Alignment Learning (TAL), the system is able to maintain a balance between classification and regression by carrying out anchor-free detection and employing TensorFlow and Keras for system optimization and performance evaluation under various metrics. The performance parameters are accuracy, loss, precision, ROC and confusion matrix.