Domain Adversarial Neural Networks (DANN) in Keras

The domain adversarial Neural Networks are vastly used for classification and recognition tasks in which the input data can be out-of-domain compared to the original training data. The algorithm, presented in this paper, is implemented in Keras. The presented Keras code allows flexible structure of the network by simply adjusting the "dann_helper" and different visualization and model manipulation such as visualization of model structure as seen in the image below: