Abstract:Visual domain adaptation aims to seek an effective transferable model for unlabeled target images by benefiting from the well-labeled source images following different distributions. Many recent efforts focus on extracting domain-invariant image representations via exploring target pseudo labels, predicted by the source classifier, to further mitigate the conditional distribution shift across domains. However, two essential factors are overlooked by most existing methods: 1) the learned transferable features should be not only domain invariant but also category discriminative; and 2) the target pseudo label is a two-edged sword to cross-domain alignment. In other words, the wrongly predicted target labels may hinder the class-wise domain matching. In this article, to address these two issues simultaneously, we propose a discriminative transfer feature and label consistency (DTLC) approach for visual domain adaptation problems, which can naturally unify cross-domain alignment with discriminative information preserved and label consistency of source and target data into one framework. To be specific, DTLC first incorporates class discriminative information by penalizing the maximum distance of data pair in the same class and the minimum distance of data pair sharing the different labels for each data into the distribution alignment of both domains. The target pseudo labels are then refined based on the label consistency within the domains. Thus, the transfer feature learning and coarse-to-fine target labels would be coupled to benefit each other in an iterative way. Comprehensive experiments on several visual cross-domain benchmarks verify that DTLC can gain remarkable margins over state-of-the-art (SOTA) nondeep visual domain adaptation methods and even be comparable to competitive deep domain adaptation ones.

Domain Class Consistency Based Transfer Learning for Image Classification Across Domains.

Discriminative Kernel Transfer Learning Via L(2,1)- Norm Minimization

Trust-aware Conditional Adversarial Domain Adaptation with Feature Norm Alignment.

Discriminative Transfer Feature and Label Consistency for Cross-Domain Image Classification.

Joint Domain Alignment and Discriminative Feature Learning for Unsupervised Deep Domain Adaptation

Domain Invariant Transfer Kernel Learning.

Transfer Kernel Sparse Coding Based on Dynamic Distribution Alignment for Image Representation

Domain Adaptation Based on the Measure of Kernel-Product Maximum Mean Discrepancy.

Heterogeneous Domain Adaptation Via Incremental Discriminative Knowledge Consistency

Adaptive Local Neighbors for Transfer Discriminative Feature Learning.

Maximal Domain Independent Representations Improve Transfer Learning

LSDT: Latent Sparse Domain Transfer Learning for Visual Adaptation

Dyadic Transfer Learning for Cross-Domain Image Classification

Feature-based Distant Domain Transfer Learning.

Guided Discrimination and Correlation Subspace Learning for Domain Adaptation

Adversarial Transfer Learning for Cross-Domain Visual Recognition

Nonconvex and discriminative transfer subspace learning for unsupervised domain adaptation

Domain Adaptation Based on Semi-Supervised Cross-Domain Mean Discriminative Analysis and Kernel Transfer Extreme Learning Machine

Domain Adaptation via Transfer Component Analysis

Domain Invariant and Class Discriminative Feature Learning for Visual Domain Adaptation

Balanced Discriminative Transfer Feature Learning for Visual Domain Adaptation