Title: Advancing automatic recognition of Moroccan Phoenix dactylifera L. varieties through transfer learning techniques

Authors: Ayoub Aoulalay; Abderrahim El Mhouti; Abdellah Elzaar; Rachida Assawab; Mohammed Massar

Addresses: ISISA, FS, Abdelmalek Essaadi University, Tetouan, Morocco ' ISISA, FS, Abdelmalek Essaadi University, Tetouan, Morocco ' Laboratory of Research and Development in Engineering Sciences (RDSI), Faculty of Sciences and Techniques Al-Hoceima, Abdelmalek Essaadi University, Tetouan, Morocco ' Laboratory of Research and Development in Engineering Sciences (RDSI), Faculty of Sciences and Techniques Al-Hoceima, Abdelmalek Essaadi University, Tetouan, Morocco ' FP, Soultan Moulay Slimane University, Khouribga, Morocco

Abstract: Date fruit holds immense significance not only commercially and in terms of health benefits but also in Moroccan cultural heritage. However, as knowledge about Moroccan date cultivars diminishes, leveraging modern technology becomes essential for both preservation and accessibility of this invaluable cultural heritage. In this study, a comprehensive approach utilising computer vision for automatic recognition of popular Moroccan date fruit cultivars is presented. Through various methodologies, including manual feature extraction and transfer learning. The dataset comprises nine cultivars, with emphasis on capturing real-world variability through image acquisition under unstable conditions. The findings indicate exceptional performance, achieving close to 99% accuracy using transfer learning.

Keywords: date fruit recognition; convolutional neural network; CNN; transfer learning; hand crafted features; deep learning.

DOI: 10.1504/IJPTI.2025.151708

International Journal of Postharvest Technology and Innovation, 2025 Vol.10 No.1, pp.60 - 82

Received: 09 May 2024
Accepted: 21 Dec 2024
Published online: 16 Feb 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article