Arecanut Classification Using MobileNetV2 and SVM: A Lightweight Deep Learning Approach for Precision Agriculture

DOI: https://doie.org/10.10399/JBSE.2025723655

Satheesha K M, Jithendra P R Nayak, Mahesh Tubaki

Keywords:

Areca nut classification, deep learning, MobileNetV2, Precision agriculture, Support Vector Machine (SVM)

Abstract:

To guarantee quality and productivity in agriculture, automated classification of Arecanuts is necessary. The time-consuming and error-prone manual examination is the backbone of traditional classification systems. Using MobileNetV2 for feature extraction as well as Support Vector Machine (SVM) for classification, this work proposes an effective deep learning model to discriminate between healthy and unhealthy Areca nuts. The model is assessed using important performance indicators after being trained on the user-defined dataset. The proposed model succeeds where more conventional CNNs have failed by striking a compromise between accuracy and computational efficiency. Outperforming traditional approaches, the experimental findings provide a 96.85% classification accuracy with 96.85% recall, 99.61% specificity, and 96.85% F1-score. The design of the model is lightweight, which allows it to be deployed in real-time on embedded and mobile devices. In order to improve agricultural decision-making, this study presents an efficient, scalable, and accurate method for automated Areca nut classification. We will be integrating with smart farming apps that use the Internet of Things (IoT) in the future.

PDF