This document presents a mammogram image classification model that uses an attention mechanism with transfer learning. It discusses image classification and common network models like VGG-16 and ResNet50. It also covers the MIAS mammogram dataset, data augmentation, the proposed model architecture using an attention layer, training parameters, and results. The conclusions are that ResNet50 performs better with an attention layer, achieving better accuracy and loss than VGG16 with or without attention. Further optimization of features and the attention layer are identified for future work.