Open-Set Hypothesis Transfer with Semantic Consistency

TL;DR

This paper proposes a novel approach for unsupervised open-set domain adaptation that does not require source data during adaptation, focusing on semantic consistency to improve classification of known and unknown classes.

Contribution

It introduces a hypothesis transfer method leveraging semantic consistency, enabling effective open-set adaptation without source data, which is a significant advancement over existing methods.

Findings

Outperforms state-of-the-art on UODA benchmarks

Effective in classifying known and unknown target data

Operates without source domain data during adaptation

Abstract

Unsupervised open-set domain adaptation (UODA) is a realistic problem where unlabeled target data contain unknown classes. Prior methods rely on the coexistence of both source and target domain data to perform domain alignment, which greatly limits their applications when source domain data are restricted due to privacy concerns. This paper addresses the challenging hypothesis transfer setting for UODA, where data from source domain are no longer available during adaptation on target domain. We introduce a method that focuses on the semantic consistency under transformation of target data, which is rarely appreciated by previous domain adaptation methods. Specifically, our model first discovers confident predictions and performs classification with pseudo-labels. Then we enforce the model to output consistent and definite predictions on semantically similar inputs. As a result, unlabeled data can be classified into discriminative classes coincided with either source classes or unknown classes. Experimental results show that our model outperforms state-of-the-art methods on UODA benchmarks.