Automated Tea Leaves Recognition: Multi-Classification Using YOLOv5 and Inception V3 Model

The quality and productivity of tea harvests around the world are greatly impacted by tea leaf diseases. To stop the spread of these diseases and put effective control measures in place, early and precise diagnosis is essential. Deep learning algorithms have become efficient devices for image classification tasks in recent years, opening up exciting possibilities for disease identification in agricultural sectors. This study focuses on identifying tea leaf conditions using the YOLOV5 and Inception v3 models. The research technique required gathering a broad dataset that included photos of both healthy and diseased tea leaves. A total number of 6556 patches in the form of different red rust, red spider, sunlight scorching, and healthy tea leaves have been extracted which have been used for training and testing purposes in the Inception V3 model. Several hyperparameters such as epochs, batch size, and momentum are initialized to achieve high performance. The YOLOV5 model achieves the highest mAP at 0.78 with a threshold value (0.5). During testing, the Inception V3 model achieves a high F1-score (94.3%) in the Red rust tea leaves classification. The outcomes showed the promise of the improved Inception v3 model as a trustworthy tool for disease identification in tea plantations and the efficiency of the model in detecting tea leaf illnesses conditions. Future applications for the multi-classification of tea leaves using the YOLOv5 and Inception V3 models are very interesting.