Early detection of cancer significantly improves the chances of successful treatment, including for oral cancer, which can be life-saving when identified in its early stages. Recently, deep learning has emerged as a powerful tool in facilitating the early diagnosis of various diseases, leading to more accurate and effective treatments. This paper introduces a deep learning fusion model for the detection and classification of oral cancer, combining the NasNetMobile and Xception architectures. A fusion-based feature extraction method is applied to train several machine learning classifiers, including Random Forest, AdaBoost, Support Vector Machine (SVM), Logistic Regression, Decision Tree, and Naive Bayes. Additionally, an Ensemble Model is developed and trained using these extracted features. The proposed framework is evaluated on the ‘Multi Cancer’ dataset, and its performance is compared with results from other state-of-the-arts techniques.