DEEP LEARNING-BASED APPROACH FOR ESTIMATING THE AGE OF PAKISTANI-GROWN RICE SEEDS

Abstract

One of the records is broadly obsessive and significant foods in the creation is rice. It plays a significant role in our everyday meals Rice is one of the most widely consumed staple foods worldwide, including in Pakistan, where its demand has significantly increased in recent years. The quality of rice, a crucial factor in its import and export, is traditionally assessed by human experts, making the process costly and inconsistent. A major aspect of rice quality assessment is distinguishing between aged and fresh rice. This study presents a Convolutional Neural Network (CNN)-based model for automatically and efficiently estimating the age of rice seeds. The dataset for this study was obtained from the Rice Research Center within the Agriculture Department in Bahawalnagar, Pakistan. A custom setup was used to capture images of rice seeds from multiple angles. A lightweight CNN model with fewer layers and parameters was developed and tested on various Pakistani-grown rice varieties, including Basmati-2000, Chenab Basmati, KSK- 133, Kissan Basmati, KSK-434, PK-1121 Aromatic, and Punjab Basmati. The proposed model demonstrated superior accuracy compared to state-of-the-art CNN models, and the dataset created is now publicly available for further research. The system has been integrated into the Agri-Tech sector as a demonstration model for automated rice age estimation.