Early Detection of Tea Leaf Pathogen using Transfer Learning Techniques

Crop diseases pose a substantial economic threat to societies, impacting plant functionality, growth, and overall agricultural activities. The widespread consumption of tea and its status as a perennial crop face numerous illnesses that threaten its yield. Early detection is crucial for effective disease management and optimal tea production. For maximum tea production and efficient disease treatment, early detection is essential. Visual inspection is a common component of traditional procedures for detecting tea sickness, although it can be laborious and prone to human mistakes. Since many of the prior systems were not very accurate, deep learning models are employed to create the best model possible for the classification of diseases of tea leaves. The suggested method for the categorization of tea leaf disease consists of three hybrid models: VGG16 with Long Short-Term Memory (LSTM), Inceptionv3 with LSTM, and NASNetMobile with LSTM. After training on an extensive dataset comprising healthy and diseased tea leaf images, the effectiveness of three hybrid models is rigorously evaluated using various performance metrics. Subsequently, the best-performing model is identified for the classification of tea leaf diseases. This paper contributes to the advancement of intelligent tea agricultural methods by showcasing the potential of Deep Learning for early disease detection, paving the way for improved disease management and enhanced tea production.