This research paper aims at investigating how the smartphone photography coupled with deep learning algorithms can close research gaps relating to inaccuracy and variation in images in diagnosing oral diseases. By using the ‘centered rule’ technique in capturing an image, this approach enables the doctors get sharp and focused images despite the fact that capturing the images may be done by other people with little knowledge in that area. A novel deep learning model was created for image classification, and the corresponding five categories of oral disorders, performed with resampling technologies to maintain diagnosis stability. From the experimental result, it also implies that rather than randomly taking the images and some preprocessing also helps to improve the deep learning model performance for the early oral cancer diagnosis. This largely tap into the ubiquity of today’s smartphone and it is highly applicable for primary health care delivery, especially in the developing world where better diagnostic equipment is rare. This paper posits that by mixing deep learning with smartphone imagery, early detection and management of oral diseases is made possible in a befitting manner that is solving global health problems in the process.