Evaluation of Deep Neural Network Domain Adaptation Techniques for Image Recognition

It has been well proved that deep networks are efficient at extracting\nfeatures from a given (source) labeled dataset. However, it is not always the\ncase that they can generalize well to other (target) datasets which very often\nhave a different underlying distribution. In this report, we evaluate four\ndifferent domain adaptation techniques for image classification tasks:\nDeepCORAL, DeepDomainConfusion, CDAN and CDAN+E. These techniques are\nunsupervised given that the target dataset dopes not carry any labels during\ntraining phase. We evaluate model performance on the office-31 dataset. A link\nto the github repository of this report can be found here:\nhttps://github.com/agrija9/Deep-Unsupervised-Domain-Adaptation.\n