Lightweight unsupervised domain adaptation by convolutional filter reconstruction

© Springer International Publishing Switzerland 2016. Recently proposed domain adaptation methods retrain the network parameters and overcome the domain shift issue to a large extent. However, this requires access to all (labeled) source data, a large amount of (unlabeled) target data, and plenty of computational resources. In this work, we propose a lightweight alternative, that allows adapting to the target domain based on a limited number of target samples in a matter of minutes. To this end, we first analyze the output of each convolutional layer from a domain adaptation perspective. Surprisingly, we find that already at the very first layer, domain shift effects pop up. We then propose a new domain adaptation method, where first layer convolutional filters that are badly affected by the domain shift are reconstructed based on less affected ones.

ISBN:978-3-319-49408-1

Publication year:2016

BOF-keylabel:yes

IOF-keylabel:yes

Authors from:Higher Education

Accessibility:Open