Unsupervised Domain Adaptation by Backpropagation

26 Sep 2014  ·  Yaroslav Ganin, Victor Lempitsky ·

Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available... Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of "deep" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation. Overall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets. read more

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Multi-target Domain Adaptation Office-31 RevGrad Accuracy 73.4 # 4
Multi-target Domain Adaptation Office-Home RevGrad Accuracy 57.9 # 3
Unsupervised Image-To-Image Translation SVNH-to-MNIST DANN Classification Accuracy 73.6% # 4

Results from Other Papers


Task Dataset Model Metric Name Metric Value Rank Source Paper Compare
Domain Adaptation HMDBfull-to-UCF RevGrad Accuracy 74.44 # 4
Domain Adaptation HMDBsmall-to-UCF TemPooling + RevGrad Accuracy 98.41 # 2
Domain Adaptation Olympic-to-HMDBsmall TemPooling + RevGrad Accuracy 90.00 # 2
Domain Adaptation UCF-to-HMDBfull RevGrad Accuracy 74.44 # 3
Domain Adaptation UCF-to-HMDBsmall TemPooling + RevGrad Accuracy 99.33 # 1
Domain Adaptation UCF-to-Olympic TemPooling + RevGrad Accuracy 98.15 # 1
Transfer Learning Office-Home DANN Accuracy 57.6 # 5

Methods


No methods listed for this paper. Add relevant methods here