1Department of Computer Science, LBS Government First Grade College, RT Nagar, Karnataka, India. Email: nazia.3182@gmail.com
2Associate Professor, Department of Computer Science and Applications, Bangalore University, Bangalore, Karnataka, India
Background: Evaluating the quality of rice seeds is vital for crop productivity, particularly in the face of changing climatic conditions and increasing food demand. Existing rice-seed classification methods either rely on complex imaging systems, handcrafted features, or lack the ability to perform both variety and age classification simultaneously. Hence, this study aims to develop a unified, deep learning-based approach that accurately classifies rice-seed varieties and age using accessible RGB imagery, for which this work proposed RiceNet-AFRI (Rice Network Attention-Fused Recurrent Integration), a spatial-temporal deep neural network trained on an open-access dataset of three rice varieties—Akitakomachi, Yandao-8, and Koshihikari.
Results: The model achieved 98.25% accuracy in variety classification and 85.9% in age classification, outperforming traditional ML methods and closely matching or exceeding state-of-the-art deep learning benchmarks.
Conclusion: RiceNet-AFRI is the first RGB-based deep learning model to offer simultaneous classification of rice-seed variety and age using an open dataset. RiceNet-AFRI offers a robust and scalable solution for rice-seed quality evaluation, with potential for future extensions into seed viability prediction.
Keywords: Rice seed classification, deep learning, spatial-temporal features, seed age prediction, RGB imaging, agricultural AI
How to cite this article: Hassan BN, Somashekara MT. Quality-Focused Classification of Rice Seed Variety and Age Using Attention-Fused Recurrent Integration. Int J Drug Deliv Technol. 2026;16(12s): 1011-1025. DOI: 10.25258/ijddt.16.12s.113
Source of support: Nil.
Conflict of interest: None