Generate To Adapt: Aligning Domains using Generative Adversarial
Networks
S. Sankaranarayanan, Y. Balaji, C. Castillo, und R. Chellappa. (2017)cite arxiv:1704.01705Comment: Accepted as spotlight talk at CVPR 2018. Code available here: https://github.com/yogeshbalaji/Generate_To_Adapt.
Zusammenfassung
Domain Adaptation is an actively researched problem in Computer Vision. In
this work, we propose an approach that leverages unsupervised data to bring the
source and target distributions closer in a learned joint feature space. We
accomplish this by inducing a symbiotic relationship between the learned
embedding and a generative adversarial network. This is in contrast to methods
which use the adversarial framework for realistic data generation and
retraining deep models with such data. We demonstrate the strength and
generality of our approach by performing experiments on three different tasks
with varying levels of difficulty: (1) Digit classification (MNIST, SVHN and
USPS datasets) (2) Object recognition using OFFICE dataset and (3) Domain
adaptation from synthetic to real data. Our method achieves state-of-the art
performance in most experimental settings and by far the only GAN-based method
that has been shown to work well across different datasets such as OFFICE and
DIGITS.
Beschreibung
Generate To Adapt: Aligning Domains using Generative Adversarial
Networks
%0 Generic
%1 sankaranarayanan2017generate
%A Sankaranarayanan, Swami
%A Balaji, Yogesh
%A Castillo, Carlos D.
%A Chellappa, Rama
%D 2017
%K GAN to_read
%T Generate To Adapt: Aligning Domains using Generative Adversarial
Networks
%U http://arxiv.org/abs/1704.01705
%X Domain Adaptation is an actively researched problem in Computer Vision. In
this work, we propose an approach that leverages unsupervised data to bring the
source and target distributions closer in a learned joint feature space. We
accomplish this by inducing a symbiotic relationship between the learned
embedding and a generative adversarial network. This is in contrast to methods
which use the adversarial framework for realistic data generation and
retraining deep models with such data. We demonstrate the strength and
generality of our approach by performing experiments on three different tasks
with varying levels of difficulty: (1) Digit classification (MNIST, SVHN and
USPS datasets) (2) Object recognition using OFFICE dataset and (3) Domain
adaptation from synthetic to real data. Our method achieves state-of-the art
performance in most experimental settings and by far the only GAN-based method
that has been shown to work well across different datasets such as OFFICE and
DIGITS.
@misc{sankaranarayanan2017generate,
abstract = {Domain Adaptation is an actively researched problem in Computer Vision. In
this work, we propose an approach that leverages unsupervised data to bring the
source and target distributions closer in a learned joint feature space. We
accomplish this by inducing a symbiotic relationship between the learned
embedding and a generative adversarial network. This is in contrast to methods
which use the adversarial framework for realistic data generation and
retraining deep models with such data. We demonstrate the strength and
generality of our approach by performing experiments on three different tasks
with varying levels of difficulty: (1) Digit classification (MNIST, SVHN and
USPS datasets) (2) Object recognition using OFFICE dataset and (3) Domain
adaptation from synthetic to real data. Our method achieves state-of-the art
performance in most experimental settings and by far the only GAN-based method
that has been shown to work well across different datasets such as OFFICE and
DIGITS.},
added-at = {2018-04-16T15:41:15.000+0200},
author = {Sankaranarayanan, Swami and Balaji, Yogesh and Castillo, Carlos D. and Chellappa, Rama},
biburl = {https://www.bibsonomy.org/bibtex/279fc54c2f1c40f3d2d295972ef8ffe6a/jk_itwm},
description = {Generate To Adapt: Aligning Domains using Generative Adversarial
Networks},
interhash = {7a8b7e17551c7c2698d7b2e788ec7437},
intrahash = {79fc54c2f1c40f3d2d295972ef8ffe6a},
keywords = {GAN to_read},
note = {cite arxiv:1704.01705Comment: Accepted as spotlight talk at CVPR 2018. Code available here: https://github.com/yogeshbalaji/Generate_To_Adapt},
timestamp = {2018-04-16T15:41:15.000+0200},
title = {Generate To Adapt: Aligning Domains using Generative Adversarial
Networks},
url = {http://arxiv.org/abs/1704.01705},
year = 2017
}