Diagnosing liver diseases using computed tomography (CT) images can be challenging even for experienced radiologists due to the complexities involved in evaluating the liver. Accurately determining the type, size and severity of tumors is often difficult. In recent years there has been a growing need for computer-assisted imaging techniques to aid in liver disease diagnosis ultimately improving clinical outcomes which in turn improves the life span of patients by early detection of the disease and treatment. This paper presents an innovative deep learning model UNet70 for liver tumor classification where CT images are categorized as either having a tumor (hepatocellular and Metastatic) or not. Our results show that the proposed model excels in terms of accuracy, sensitivity and dice score compared to other established algorithms and demonstrates excellent adaptability across various datasets. With an accuracy of 94.58 %, dice score of 94.73 % and sensitivity of 97.50 % the model outperforms existing methods showcasing its effectiveness.