This project is an application of two deep learning models -- ResNets and DenseNets, to identify breast cancer using MRI images. I use a public dataset on Kaggle to train 3 ResNets and 3 DenseNets with different number of layers. ResNet-152V2 and DenseNet-169 achieve the highest validation accuracy of 99.65%, but DenseNet-169 has better parameter efficiency than ResNet-152V2.
Please see 680Project.pdf for the project report.
Implementation details are in CS680project.ipynb.