In this assignment you are expected to build, train and evaluate a deep learning framework on a standard setting of Unsupervised Domain Adaptation (UDA). As you have studied during the theoretical course, UDA is a generic expression that covers any possible technique designed to address the issue of the domain shift that may exist between two or more data distributions.
In particular, for this assignment you will be provided with a UDA benchmark consisting of two datasets. You will be required to treat one of the two datasets as source domain and the other as target domain, and to propose a UDA technique to counteract the negative impact of the domain gap when training your model on the source distribution and evaluating it on the target distribution.
As it holds for any standard UDA framework, the quality of the domain alignment strategy shall essentially be assessed by looking at the gain obtained with your proposed framework over the so called source-only baseline.