More than half of the world population, mostly in Asia, Africa and Latin America, relies on rice as a food source, which ensures food security and finances, mostly due to its genetic purity, physical characteristics, and its viability status, as it directly influences yields, food security, and market price, but the industry is planked by fraud and adulteration. Manual classification methods are time-consuming, subjective, and unreliable, and an automated response to consistency and scale requirements is needed. In addition, the proposed study presents a new dataset of 1,600 images of eight varieties of rice in Bangladesh that were taken in the fields and enhanced to 3,200 using SMOTE, which combats data scarcity. With the latest CNN technology, DenseNet201 has reached an accuracy of 96.25% with a 0.1816 loss that is worse than some CNNs, such as Xception and InceptionV3. SHAP analysis allowed more interpretability and showed the main properties of seed shape and texture. This system helps in the automation of quality control as well as fighting market fraud, which is useful to farmers and players in the Bangladesh Agricultural sector. The paper presents future directions, results, and methodology.