Imbalanced Open Set Domain Adaptation via Moving-threshold Estimation and Gradual Alignment
This addresses a domain-specific problem for multimedia applications involving cross-domain knowledge transfer, but it is incremental as it builds on existing OSDA methods by adding handling for label shift and class imbalance.
The paper tackles the problem of imbalanced open set domain adaptation (IOSDA), where covariate shift, label shift, and category mismatch occur simultaneously, by proposing the OMEGA method to improve performance on class-imbalanced data. The result shows that OMEGA significantly outperforms existing state-of-the-art methods on IOSDA, OSDA, and OPDA benchmarks.
Multimedia applications are often associated with cross-domain knowledge transfer, where Unsupervised Domain Adaptation (UDA) can be used to reduce the domain shifts. Open Set Domain Adaptation (OSDA) aims to transfer knowledge from a well-labeled source domain to an unlabeled target domain under the assumption that the target domain contains unknown classes. Existing OSDA methods consistently lay stress on the covariate shift, ignoring the potential label shift problem. The performance of OSDA methods degrades drastically under intra-domain class imbalance and inter-domain label shift. However, little attention has been paid to this issue in the community. In this paper, the Imbalanced Open Set Domain Adaptation (IOSDA) is explored where the covariate shift, label shift and category mismatch exist simultaneously. To alleviate the negative effects raised by label shift in OSDA, we propose Open-set Moving-threshold Estimation and Gradual Alignment (OMEGA) - a novel architecture that improves existing OSDA methods on class-imbalanced data. Specifically, a novel unknown-aware target clustering scheme is proposed to form tight clusters in the target domain to reduce the negative effects of label shift and intra-domain class imbalance. Furthermore, moving-threshold estimation is designed to generate specific thresholds for each target sample rather than using one for all. Extensive experiments on IOSDA, OSDA and OPDA benchmarks demonstrate that our method could significantly outperform existing state-of-the-arts. Code and data are available at https://github.com/mendicant04/OMEGA.