Deep Learning-Based Tea Fermentation Grading

Tea is the most frequently drunk beverages all over the world. There are more than 15,000 different kinds of tea based on processing methods, but the most common ones are oolong, yellow, ilex, matcha, black, sencha, green teas, and others. The uttermost widely consumed type of tea is black among the global categories. The black tea can be prepared by plucking, cutting, withering, curling, tearing, fermenting, sorting, and drying are complete steps in the preparation of black tea. Although each of these steps has an impact on the processed quality of tea, the utmost insistent is the fermentation because it determines the quality directly. As the process is time-bounded, fermentation is now manually judged by tea tasters who watch for colour changes, smell the tea, and taste it as the fermentation process moves along. The ideal fermentation of black tea is investigated in this research using the IoT, image processing, and DCNN using maximum voting approaches. The deep learner had a faultless precision and accuracy of 1.0 each when tested on stored data. When dataset of real time is tested, the deep learning reports the highest accuracy and precision, i.e. 0.9589 and 0.8646, respectively. Additionally, when a maximum voting approach was used in decision, the deep learner obtained accuracy and precision of 0.8953 and 0.9737, respectively. It is clear from the data that the model may be benefitted to track the types of fermenting tea, incorporates black and oolong tea. The model can be expanded by being retrained to monitor the crops fermentation, such as cocoa and coffee.