Automatic Mass Detection and Localization in Mammogram using advanced Deep CV Models

This paper centers its attention on classification and localization of malignant masses in mammograms. The novel advanced neural network approach with Non-Maximum Suppression, finds the overlapping entities that detects multiple breast masses. With standard criterion the best boundary has been chosen that has probability threshold. The model gave significant contribution to the field of mammogram analysis and offers a promising avenue for improving accuracy of malignant mass detection and classification with a bounding box. The implemented model obtained the accuracy percentage of 97.2 at 0.3 on mias to identify breast tumor with a bounding box and best Intersection of Union.