Machine Learning Pipelines for Image-Based Classification of Rice Grain Varieties: A Systematic Literature Review

Rice is one of the staple foods whose value varies depending on the varieties. Inspection for its authenticity and purity is a standard process in the supply chain. Over the past decade, image-based machine learning has increasingly been studied as a non-destructive approach for rice variety classification using grain visual information. To methodically portray the existing work, this study conducts a systematic literature review by gathering, summarizing, and grouping related studies into categories based on their workflows. From 101 studies, the identified pipelines were categorized as classification-only, separation-and-classification, and reconstruction-and-classification structures. The review summarizes study characteristics, input transformations, model categories, reported evaluation metrics, normalized performance summaries, and cross-structure findings, with indexed end-to-end tables provided for further reference. Overall, this review supports researchers in selecting workflow designs aligned with their objectives or deployment conditions, while highlighting limitations, research directions, and reporting recommendations to improve reproducibility and cross-study comparability in future studies.