PLANT SPECIES IDENTIFICATION USING MACHINE LEARNING

Accurate identification of plant species plays a crucial role in agriculture, environmental monitoring, andbiodiversity conservation. Traditional methods of species recognition rely heavily on expert knowledge, whichis often time‑consuming and prone to human error. To address these challenges, this study presents a deeplearning–based approach for plant species identification using convolutional neural networks (CNNs) andtransfer learning. A custom dataset of 48 plant species was utilized, with extensive preprocessing andaugmentation techniques applied to improve generalization. Multiple state‑of‑the‑art models, includingEfficientNet‑B0, MobileNetV3‑Large, and RexNet‑150, were trained and evaluated using a standardized pipelinebuilt on PyTorch and the timm library. Model performance was assessed through accuracy, F1‑score, andconfusion matrix analysis, ensuring robust evaluation across all classes. To enhance usability, a graphical userinterface (GUI) was developed with Tkinter, enabling non‑technical users to classify plant leaves and viewdetailed prediction reports. Experimental results demonstrate that the proposed framework achieves highclassification accuracy while maintaining computational efficiency, with EfficientNet and RexNet showingparticularly strong performance. This research highlights the effectiveness of transfer learning–based CNNs inplant identification tasks and offers a scalable, user‑friendly solution for real‑world applications in agricultureand ecology.Key Words - Melanoma Detection, Vision Transformer (ViT), Transfer Learning, EfficientNet,Dermoscopic Image Analysis.