Stretching Domain Adaptation: How far is too far?

  • 2017-12-06 17:03:07
  • Yunhan Zhao, Haider Ali, Rene Vidal
  • 7

Abstract

While deep learning has led to significant advances in visual recognitionover the past few years, such advances often require a lot of annotated data.While unsupervised domain adaptation has emerged as an alternative approachthat doesn't require as much annotated data, prior evaluations of domainadaptation have been limited to relatively simple datasets. This work pushesthe state of the art in unsupervised domain adaptation through an in depthevaluation of AlexNet, DenseNet and Residual Transfer Networks (RTN) onmultimodal benchmark datasets that shows and identifies which layers moreeffectively transfer features across different domains. We also modify theexisting RTN architecture and propose a novel domain adaptation architecturecalled "Deep MagNet" that combines Deep Convolutional Blocks with multipleMaximum Mean Discrepancy losses. Our experiments show quantitative andqualitative improvements in performance of our method on benchmarking datasetsfor complex data domains.

 

Introduction (beta)

None

 

Conclusion (beta)

None