A malignant tumor known as breast cancer is caused by the rapid and unchecked growth of cells in the breast tissue. Manual interpretation is frequently used in traditional diagnostic techniques, which laborious, subjective, and error-prone. This research proposes a sophisticated Deep Learning [DL] vision for automated identification and precise prediction of breast cancer from diagnostic imaging modalities in order to get over these drawbacks. Initially to improve image quality and reduce noise, this method starts with preprocessing with a bilateral filter. Segmentation is then performed using a level set technique to isolate ambiguous regions. Superpixel-based region features are extracted to capture local texture and structural details. For efficient categorization, these features are supplied into a deep vision neural network. The deep network has been trained to make highly accurate distinctions between benign and malignant tissues. This system achieves an outstanding performance with an accuracy, precision, recall, and F1-score of 96% and it is implemented using Python programming language.