Boosting Domain Adaptation by Discovering Latent Domains
This work addresses a key limitation in domain adaptation for computer vision, where datasets often have hidden domain structures, making it significant for researchers and practitioners in visual recognition tasks.
The paper tackles the problem of domain adaptation when source data contains multiple latent domains, which existing methods often assume is a single distribution, leading to sub-optimal performance. It introduces a CNN architecture that automatically discovers these latent domains and uses this information to align feature distributions, resulting in large-margin improvements over state-of-the-art multi-source DA methods on public datasets.
Current Domain Adaptation (DA) methods based on deep architectures assume that the source samples arise from a single distribution. However, in practice, most datasets can be regarded as mixtures of multiple domains. In these cases exploiting single-source DA methods for learning target classifiers may lead to sub-optimal, if not poor, results. In addition, in many applications it is difficult to manually provide the domain labels for all source data points, i.e. latent domains should be automatically discovered. This paper introduces a novel Convolutional Neural Network (CNN) architecture which (i) automatically discovers latent domains in visual datasets and (ii) exploits this information to learn robust target classifiers. Our approach is based on the introduction of two main components, which can be embedded into any existing CNN architecture: (i) a side branch that automatically computes the assignment of a source sample to a latent domain and (ii) novel layers that exploit domain membership information to appropriately align the distribution of the CNN internal feature representations to a reference distribution. We test our approach on publicly-available datasets, showing that it outperforms state-of-the-art multi-source DA methods by a large margin.