Breast cancer (BC) is one of the most common diseases in women globally, according to publications from the world health organization and the global cancer observatory. Nearly 48% of the world's 130,000 women have breast cancer. Additionally, 39% of women with cancer had BC. Early and accurate detection of breast cancer remains a significant challenge due to the complex nature of mammographic images and the limitations of traditional diagnostic approaches. This study utilizes the CoroNet model based on the Xception architecture, trained on the CBIS-DDSM dataset using ImageNet weights, to classify mammograms into multiple diagnostic categories. A CNN method called CoroNet is suggested for use in automatically detecting BC using the CBIS-DDSM dataset. Built on whole-image BC derived from mammograms and trained on ImageNet data, the Xception architecture forms its backbone. This study makes use of the convolutional design approach because of its superior performance compared to the other methods. Both the training and testing of CoroNet was placed on the supplied dataset. Results from experiments suggest that it is possible to achieve an overall accuracy of 95% when classifying benign masses as opposed to malignant masses and benign calcifications as opposed to malignant calcifications in a four-class classification. In the two-category problem, where calcifications and masses are both classified, CoroNet achieves an accuracy of 89%. The study found that with additional training data available, promising results may be much better.