Efficient Rice Grain Classification Using MobileNetV2: A Deep Learning Approach

Rice grain categorization guarantees quality control and variety identification needed for commerce and consumption, therefore serving a critical function in agriculture. Five rice varieties: Arborio, Basmati, Ipsala, Jasmine, and Karacadag are categorized in this work using the MobileNetV2 model. High-resolution photos displaying these deviations collected from the Kaggle open source website make up the dataset. Our approach pre-processes the photos, uses transfer learning to train the MobileNetV2 model, and then assesses its performance using conventional benchmarks. Across all classes, the model produced outstanding performance showing great accuracy, recall, and F1-scores. The model especially generated a precision of 0.99, recall of 0.99, and F1-score of 0.99 for Arborio, Basmati, Jasmine, and Karacadag who scored a perfect 1.00 in all metrics Ipsala. Reflecting typically robustness and efficiency, the model had an accuracy of 0.99. These findings show that MobileNetV2 is quite capable of differentiating between several rice varieties, therefore providing a dependable and effective instrument for automated rice grain categorization. The results of this study have major consequences for enhancing quality control and automation in rice farming, therefore enabling more accurate and effective sorting techniques.