Oral cancer stances a significant world health challenge, with new cases and deaths reported worldwide. Despite advancements in medical science, the disease remains a formidable adversary, ranking as the 13th most prevalent cancer globally. This paper explores how convolutional neural networks (CNNs) can be utilized to detect and classify oral cancer from medical imaging data. Leveraging transfer learning techniques, we employ a ResNet-based architecture to capitalize on pre-trained weights obtained from a general image recognition task. A dataset is composed of histopathological images of oral lesions annotated with benign or malignant labels in a meticulous way. Data augmentation techniques are employed to diversify the dataset so as to ensure robust model training and evaluation. The ResNet model that has been trained using transfer learning shows remarkable validation accuracy (96.72%) and low validation loss (0.1712) indicating its ability to uncover hidden knowledge into simplicity. Further, the said model also boasts of high training accuracy (98.25%) and low training loss (0.1501) thus demonstrating its efficiency on learning through training examples’. Such findings demonstrate the promise of CNNs specifically ResNet with transfer learning toward enhancing oral cancer detection hence leading to better diagnostic accuracy as well as patient outcomes.