Breast cancer has turn into an inevitable cause of women mortality in developed, developing and underdeveloped countries. The early detection as well as accurate classification of this illness enables the medical practitioners to take timely decisions on appropriate medical interventions therapeutic options. Digital mammography is a broadly accepted medical imaging technique, used for early screening, detection, and diagnosis. This research was to create a refined deep learning model, termed as FTResNet50V2, developed using ResNet50V2 architecture. This system is exclusively tailored for breast cancer detection from digital mammography using DDSM dataset. The hyperparameter tuning was performed with a sequence of tests conducted to evolve a robust model. The performance of the FTResNet50V2 model was enhanced through the ablation studies. The effectiveness of the proposed FTResNet50V2 on the breast cancer classification was measured using accuracy, precision, recall, and F1-score. The FTResNet50V2 model is proven to produce outstanding results.