2025 VOLUME 5 ISSUE 1
RESEARCH ARTICLE
Enhancing precision in banana variety classification using deep learning model
AUTHOR
Md Ripon Sheikh, Md. Masudul Islam, Galib Muhammad Shahriar Himel
ABSTRACT
This research presents a comprehensive approach to classifying six banana varieties commonly found in Bangladesh using a customized transfer-learning model. The varieties include Shagor, Shabri, Champa, Anaji, Deshi, and Bichi. We developed a comprehensive banana dataset comprising 6,000 high-resolution images, captured under diverse conditions to ensure robust model training and testing. A model is proposed to enhance feature extraction and classification performance, incorporating additional layers and fine-tuning for the specific characteristics of the banana varieties. The proposed model achieved an outstanding accuracy of 99.67%, surpassing the performance of existing state-of-the-art methods. This result demonstrates the model's exceptional ability to distinguish between similar banana varieties with high precision, even in the presence of real-world background variations. It is a valuable tool for agricultural applications and large-scale supply chain management, enhancing automation and quality assurance in banana processing industries. The dataset and the high accuracy achieved in this study set a new benchmark for banana classification tasks, highlighting the potential of deep learning techniques in improving the efficiency and accuracy of crop identification processes. This research contributes to the growing knowledge of applying advanced machine learning models to agricultural problems, with significant implications for food security and sustainable farming practices.
KEYWORDS
Banana variety classification; Fruit classification; Deep learning; Agricultural applications
