Learning to Generate the Unknowns for Open-set Domain Adaptation

In many situations, the data one has access to at test time follows a different distribution from the training data. Over the years, this problem has been tackled by closed-set domain adaptation techniques. Recently, open-set domain adaptation has emerged to address the more realistic scenario where additional unknown classes are present in the target data. In this setting, existing techniques focus on the challenging task of isolating the unknown target samples, so as to avoid the negative transfer resulting from aligning the source feature distributions with the broader target one that encompasses the additional unknown classes. Here, we propose a simpler and more effective solution consisting of complementing the source data distribution and making it comparable to the target one by enabling the model to generate source samples corresponding to the unknown target classes. To this end, we attach a generative model to a standard domain adaptation network and augment the source data with the generated samples before matching the source distribution to the target one, thus avoiding negative transfer between the domains. We formulate this as a general module that can be incorporated into any existing closed-set approach and show that this strategy allows us to outperform the state of the art on standard open-set domain adaptation benchmark datasets.