Overcoming dataset bias: An unsupervised domain adaptation approach

Recent studies have shown that recognition datasets are biased. Paying no heed to those biases, learning algorithms often result in classifiers with poor cross-dataset generalization. We are developing domain adaptation techniques to over-come those biases and yield classifiers with significantly improved performance when generalized to new testing datasets. Our work enables us to continue to har-vest the benefits of existing vision datasets for the time being. Moreover, it also sheds insights about how to construct new ones. In particular, domain adaptation raises the bar for collecting data — the most informative data are those which cannot be classified well by learning algorithms that adapt from existing datasets.