The identification of medicinal plants plays a critical role in healthcare, particularly in traditional medicine systems. Manual identification methods are often inaccurate, time-consuming, and require expert knowledge. Recent advances in deep learning have enabled automated plant recognition using image-based techniques. This paper presents a comparative study between an existing cascaded deep learning model utilizing Particle Swarm Optimization (PSO) and a proposed hybrid system based on Convolutional Neural Networks (CNN) enhanced with PSO and integrated Ayurvedic knowledge.

The existing model employs a combination of pre-trained CNN, PSO-based feature selection, and Support Vector Machine (SVM) classification to achieve high accuracy. However, it lacks interpretability and practical usability. The proposed system addresses these limitations by combining CNN-based classification with PSO-driven optimization and an additional knowledge layer that provides medicinal insights. Experimental evaluation demonstrates that while both systems achieve high accuracy, the proposed system offers improved usability and real-world applicability.

The study highlights the importance of integrating domain knowledge with intelligent models for effective healthcare solutions.