Abstract: This study aimed to develop a herbal plant classification system
based on Deep Learning technology implemented on a mobile application, and
to evaluate both the system’s performance and user satisfaction. The application
was developed using the Flutter framework for mobile development, PHP for
data management, and TensorFlow for constructing the Deep Learning model.
A total of 42 medicinal plants were collected, comprising 2,100 images. The
system employed a Custom CNN Architecture consisting of four convolutional
layers, and its performance was tested using the Adam optimizer. The results
demonstrate that the herbal plant classification system achieved an accuracy of
85.7Percent. In addition, experts evaluated the system’s performance at the highest
level (𝑥̅ = 4.28, S.D. = 0.14), and users reported the highest level of satisfaction
(𝑥̅ = 4.21, S.D. = 0.61). These findings indicate that the developed classification
system is feasible for real-world use and can contribute significantly to the
preservation of medicinal plants.