Oral cancer, prevalent in the head and neck region and primarily associated with tobacco and excessive alcohol consumption, often progresses insidiously, reaching advanced stages before detection. This paper introduces a novel approach presenting a customized Convolutional Neural Network (CNN) model for the early detection and classification of oral cancer using 2D smartphone mouth images. This cost-efficient method aims to be accessible to a broader population. The customized CNN model incorporates an increased number of filters to effectively extract features crucial for detecting cancerous lesions. Demonstrating exceptional performance, the model achieves an impressive accuracy of 96.31%, showcasing its efficiency in oral cancer detection. This paper represents a significant step toward autonomous and accessible oral cancer diagnosis. Expanding the scope of research to encompass a larger dataset could significantly enhance the efficacy of oral cancer detection, surpassing the limitations posed by the dataset utilized in this study and yielding more robust outcomes.