An Image Dataset for Automated Identification and Rice Quality Analysis

Rice is one of the most widely cultivated and essential crops in Bangladesh. However, there are wide varieties of rice, and the rice grains are so small that not only general people but also the local sellers often cannot distinguish the variety or quality of rice, which causes a scam to the buyers. To tackle this challange we have introduced a raw photobased dataset of fifteen Bangladeshi rice varieties. They are BRRI Ataish, Durga Ataish, Aman(Less fibre), Manikganj Aman(Full Fibre), Bangla Moti(Kushtiya), Biroi(Mymensingh), Black Rice(Sylhet), Guti Shorna(Dinajpur), Haski 29, Biroi(Sylhet), Lal Gutishorna, Miniket, Nazirshail, Paijam, and Utshob Nazir are captured under diverse conditions.
The dataset contains a total of 4922 high-resolution images. To ensure robustness and diversity, all images were originally captured in PRO format using Samsung Galaxy S24 mobile phone cameras under different lighting environments. After capturing, these images were converted into JPG format, and the size was 3060*4080 to maintain all the image quality.
This rice grain image database is designed to support research in machine learning and computer vision, particularly for automated rice type identification and quality analysis.For upcoming deep learning models in rice identification, quality assurance, and enhanced market transparency across Bangladesh, it will serve as a standard.