Early diagnosis of diabetes is essential, as it makes it possible to reduce and prevent vascular problems for the patient. This study aimed to develop an efficient computer vision model for diagnosing Type 2 diabetes mellitus (T2DM) using artificial intelligence and machine learning techniques. The proposed approach is tested with various architectures yielding accuracy levels exceeding 92%, with the best models achieving over 95%, peaking at 99.58% using a proprietary architecture and support vector machine (SVM) classifier. Preprocessing tasks, including data normalization and augmentation, are consistently applied during implementation. The proprietary architecture demonstrated an accuracy of 99.3%, showcasing the efficacy of convolutional neural networks (CNN) integrated with SVM classifier further improved performance to 99.6%, with comparable solution times. Future research will explore alternative SVM formulations. From the experimental results it is stated that the proposed method underscores the effectiveness of CNN and SVM techniques in diagnosing T2DM and suggests promising implications for further investigation.