Contrastive Adaptation Network for Unsupervised Domain Adaptation
About
Unsupervised Domain Adaptation (UDA) makes predictions for the target domain data while manual annotations are only available in the source domain. Previous methods minimize the domain discrepancy neglecting the class information, which may lead to misalignment and poor generalization performance. To address this issue, this paper proposes Contrastive Adaptation Network (CAN) optimizing a new metric which explicitly models the intra-class domain discrepancy and the inter-class domain discrepancy. We design an alternating update strategy for training CAN in an end-to-end manner. Experiments on two real-world benchmarks Office-31 and VisDA-2017 demonstrate that CAN performs favorably against the state-of-the-art methods and produces more discriminative features.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Unsupervised Domain Adaptation | Office-Home (test) | Average Accuracy68.9 | 332 | |
| Image Classification | Office-31 | Average Accuracy90.6 | 261 | |
| Domain Adaptation | Office-31 unsupervised adaptation standard | Accuracy (A to W)94.5 | 162 | |
| Domain Adaptation | Office-31 | Accuracy (A -> W)94.5 | 156 | |
| Unsupervised Domain Adaptation | ImageCLEF-DA | Average Accuracy89.6 | 104 | |
| Image Classification | Office-31 (test) | Avg Accuracy90.6 | 93 | |
| Object Classification | VisDA synthetic-to-real 2017 | Mean Accuracy87.2 | 91 | |
| Unsupervised Domain Adaptation | VisDA unsupervised domain adaptation 2017 | Mean Accuracy87.2 | 87 | |
| Image Classification | VisDA-C (test) | Mean Accuracy87.2 | 76 | |
| Domain Adaptation | Image-CLEF DA (test) | Average Accuracy89.6 | 76 |