Breast cancer detection is crucial for reducing mortality and improving survival rates in women. Convolutional neural networks (CNNs) have demonstrated high accuracy, approximately 95%, in detecting breast cancer in mammograms and magnetic resonance imaging (MRI), which facilitates effective treatments and enhances patient recovery. In this project, we developed a breast cancer detection system using DenseNet169 and a custom CNN architecture. The system employed Gaussian and bilateral filters to enhance image contrast and edges, reducing noise and improving the visualization of tumors and breast tissue. Our model achieved classification accuracy of 63% for distinguishing between benign and malignant images, with training and validation accuracies of 95% and 93%, respectively. This indicates the model’s robustness in learning from the dataset with minimal overfitting. Compared to other CNN architectures like AlexNet and ResNet50, our DenseNet-based model required fewer training epochs while maintaining high accuracy, demonstrating the efficiency of the filter-augmented approach. These results suggest significant potential for early breast cancer detection, which could lead to improved patient outcomes and higher survival rates.