 The current study evaluated two deep learning models, YOLO-V5X and YOLO-V5M, to detect carious lesions from smartphone images. Both models demonstrated high accuracy in identifying carious lesions, with YOLO-V5X achieving the highest accuracy of 0.727. This model also had the best precision and recall scores, indicating its ability to accurately identify carious lesions. Additionally, no single model was found to be suitable for diagnosing all types of carious lesions, suggesting that multiple models may need to be used together to achieve optimal diagnostic accuracy. This article was authored by S. M. Syama Salahin, M. D. Shifatullah, Saif Ahmed, and others.