Orbital tumors are the most common type of tumor affecting the orbit. Some factors, such as technical causes relating to imaging quality and human error, contribute to radiologists misdiagnosing eye tumors. Computer-aided detection systems (CADs) are being developed to address these limitations and have recently been used in numerous imaging modalities for eye tumor diagnosis. CAD technologies increase radiologists' ability to detect and distinguish between normal and diseased tissues. These techniques are only conducted as a second opinion, but the radiologist makes the final decisions. This article presents the contemporary CAD method for detecting orbital tumors on magnetic resonance imaging (MRI) utilizing Convolutional Neural Networks (CNN). Pre-processing, Data Augmentation, Classification, and Evaluation are the four stages that involve our CAD system. Two datasets were used for MRI images: 1404 MRI T1-weighted images and 1560 MRI T2-weighted images. The system was evaluated by many evaluation metrics including the recognition rate which gives 95% for T1-weighted images and 94% for T2-weighted images.