Abstract:
"Ensuring the quality of tea powder in the production industry is a critical facet of maintaining
industry standards. This research introduces an innovative machine learning-based system
dedicated to the precise assessment of tea quality. By amalgamating image processing
techniques, Convolutional Neural Networks (CNN) and VGG16 Pre-trained model, the system
demonstrates efficiency in categorizing tea powder into distinct grades.
The study is underpinned by a diverse dataset comprising 5000 images of tea powder across
five unique classes. Employing a comprehensive workflow that encompasses image preprocessing and feature extraction, model achieves an impressive accuracy rate of 95.7% in
contumely created CNN model (self-composed) and 94% in the model created based on
VGG16 model. The architectural design of the system facilitates the seamless integration of
image acquisition, processing, and classification, contributing to its overall effectiveness.
Results gleaned from this research underscore the system's proficiency in distinguishing
between different grades of tea powder, marking it as a valuable tool for tea manufacturers
seeking to ensure quality standards.
Due to the highest accuracy performance of contumely created CNN model that choose to
create the web based application of tea leaf recognising and classification. This research not
only addresses the pressing need for efficient and accurate grading in the tea production process
but also holds promise for broader applications in automated quality assessment across various
industries. As the demand for consistent and high-quality products continues to rise, the
presented system stands as a testament to the potential of advanced technologies in meeting
these evolving needs."