EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Deep Learning-Based Classification of Malaysian Pineapple Ripeness Index Using Fine-Tuned VGG16

Maslina Abdul Aziz, Ida Normaya Mohd Nasir, Nurul Ain Mustakim, Shuzlina Abdul Rahman, Jemal Abawajy

International Journal on Advanced Science, Engineering and Information Technology · 20.06.2026 · 16 (3) · 818-823

DOI: 10.18517/ijaseit.16.3.21488 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal on Advanced Science, Engineering and Information Technology, online erschienen am 20.06.2026. Verfasst von Maslina Abdul Aziz und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Pineapple ripeness grading is a critical postharvest operation; however, the Malaysian industry still relies heavily on subjective visual inspection, which leads to significant inconsistencies in quality control. This study bridges the gap between traditional manual practices and the rigorous demands of automated export standards by developing a deep learning framework centered on the VGG16 architecture. The research focuses specifically on the Malaysian MD2 pineapple, with the dataset meticulously labeled according to the Malaysian Pineapple Industry Board (MPIB) 7-stage ripeness index. Two distinct modeling strategies were evaluated: training the network from scratch with random weight initialization and employing transfer learning with pre-trained ImageNet weights. Following structured pre-processing and data augmentation to enhance dataset diversity, the models were validated across 90:10, 80:20, and 70:30 training-validation splits. The results demonstrate that the transfer learning approach significantly outperformed scratch-training, achieving a peak accuracy of 98% and a minimal loss of 5%. Beyond its high precision, the transfer learning model exhibited faster convergence and greater stability across all ripeness categories. To translate these academic findings into a practical field tool, the optimal model was deployed as a real-time web application using the Streamlit framework. This deployment provides stakeholders with an objective, data-driven tool for instant ripeness prediction, effectively mitigating human bias in the sorting pipeline. By aligning computational performance with national grading standards, this system offers a scalable solution to maintain the quality required for Malaysia’s competitive MD2 export market.

Zitieren

APA

Aziz, M. A., Nasir, I. N. M., Mustakim, N. A., Rahman, S. A., & Abawajy, J. (2026). Deep Learning-Based Classification of Malaysian Pineapple Ripeness Index Using Fine-Tuned VGG16. International Journal on Advanced Science, Engineering and Information Technology, 16(3), 818-823. https://doi.org/10.18517/ijaseit.16.3.21488

Vancouver

Aziz MA, Nasir INM, Mustakim NA, Rahman SA, Abawajy J. Deep Learning-Based Classification of Malaysian Pineapple Ripeness Index Using Fine-Tuned VGG16. International Journal on Advanced Science, Engineering and Information Technology. 2026;16(3):818-823. doi:10.18517/ijaseit.16.3.21488

BibTeX
@article{Aziz2026,
  author = {Aziz, Maslina Abdul and Nasir, Ida Normaya Mohd and Mustakim, Nurul Ain and Rahman, Shuzlina Abdul and Abawajy, Jemal},
  title = {Deep Learning-Based Classification of Malaysian Pineapple Ripeness Index Using Fine-Tuned VGG16},
  journal = {International Journal on Advanced Science, Engineering and Information Technology},
  year = {2026},
  volume = {16},
  number = {3},
  pages = {818--823},
  doi = {10.18517/ijaseit.16.3.21488}
}
Kostenlos starten →