Dataset Amount and Augmentation Process Effect for CNN Rice Seed Classification

Rice seed classification is one of the critical procedures in quality assessment in the agricultural sector, and it can be made easier using machine learning. The classification process using machine learning starts with creating a dataset, preprocessing the data, and then creating and training the model. In the process of preprocessing data, there are times when augmentation is needed to process it before it is trained on the machine learning model that has been created. In this study, the author compares the results of rice seed classification for superior and non-superior classes between previously augmented data and data without augmentation. Another variation was the number of datasets used to see differences in classification results. This research shows that models using augmented data produce higher accuracy and precision than those without augmentation. In this research, the percentage increase in accuracy, precision, recall, and f1 score before and after the data was augmented was 9% to 15%. This research contributes to developing a rice seed identification system, which can increase efficiency and accuracy in the seed selection process.