Deep Learning-Based Plant Identification Using Mobile Net and Res Net

Abstract

Deep Learning-Based Plant Identification Using Mobile Net and Res Net for identifying medicinal plants. Bohera, Devilback bone, Haritoki, Lemongrass, Nayontara, Neem, Pathaorkuchi, Thankuni, Tulsi, and Zenora are the ten classes of medicinal plants that make up the dataset. In order to improve model generalization, ImageDataGenerator is used for image preprocessing and augmentation. The Adam optimizer is used to optimize both models after they have been compiled with categorical cross-entropy loss. Accuracy measurement is part of performance evaluation, and it is shown using bar charts and graphs. While both MobileNetV2 and ResNet50 achieve high accuracy, a comparison of the two models shows that MobileNetV2's lightweight architecture makes it more effective for real-time applications.
The web-based interface of the system was created with HTML, CSS, and JavaScript. Users can upload images to the frontend, and the Flask-implemented backend processes the image and returns the predicted medicinal plant class and confidence score. Through smooth user interaction made possible by this integration, medicinal plant identification is now quick and easy. By automating plant identification, this research provides a useful tool for conservation, agriculture, and healthcare. The suggested system offers a precise and effective way to identify medicinal plants by utilizing deep learning. Future research can look into improving model robustness and incorporating more plant species.