Dentists, especially those who are not oral lesion specialists and live in rural areas, need an artificial intelligence (AI) system for accurately assisting them in screening for oral cancer that may appear in smartphone images. Not many literatures present a viable model that addresses the needs, especially in the context of oral lesion segmentation in smartphone images. This study demonstrates the use of a deep learning-based AI for simultaneously identifying types of oral cancer lesions as well as precisely outlining the boundary of the lesions in the images for the first time. The lesions of interest were oral potentially malignant disorders (OPMDs) and oral squamous cell carcinoma (OSCC) lesions. The model could successfully (1) detect if the images contained the oral lesions, (2) determine types of the lesions, and (3) precisely outline the boundary of the lesions. With future success of our project, patients will be diagnosed and treated early before the pre-cancer lesions can progress into deadly cancerous ones.