Revolutionizing Tea Leaf Disease Detection for Sustainable Cultivation using Convolutional Neural Network

The tea production industry in India plays an essential role in the agricultural sector and makes a significant contribution to the overall economy of the country. India has become internationally accepted as the primary consumer and the second-highest producer of tea. This field offers employment opportunities, both directly and indirectly, to a significant number of people, particularly those living in rural regions, where it serves as their primary means of making income. The recent surge in deep learning's popularity has led to a significant increase in scientific interest in image recognition technologies, especially in areas like computerized image classification and the detection of plant diseases. India annually generates a significant amount of tea, and this has had a beneficial effect on the country's economic situation. Nevertheless, the tea industry is faced with a significant challenge due to the enormous accumulation of tea waste resulting from tea leaf diseases. This issue must be dealt with successfully to ensure the sustained prosperity of this sector. The growth rate of tea leaves can be influenced by diseases, which in turn might have an impact on the overall production of the crop. Hence, the primary objective of this research is to utilize a convolutional neural network (CNN) model to identify infected tea plants. This model is fine-tuned by setting its epoch values to 10, 20, 30, and 40, which resulted in a 99% training accuracy and 89% testing accuracy has been achieved at epoch 40. This has been resulted by utilizing a dataset consisting of many different kinds of tea leaf images. CNN is a highly appropriate model for this particular application due to its capability to quickly evaluate various types of diseases in leaves, including tea leaves and other plant leaves, hence enhancing the exactness of the findings. Additionally, the approach under examination has analyzed images of tea leaves and has identified a validation loss of 0.76. Additionally, in subsequent versions, the images can be enhanced even more by the application of augmentation approaches, hence delivering more favorable outcomes.