Generate to Adapt: Aligning Domains Using Generative Adversarial Networks

By S. Sankaranarayanan, Yogesh Kumar Balaji, Carlos Domingo Castillo, Ramalingam Chellappa. 2017

680 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Generative Models

In Representation & Self-Supervised Learning

Citations in the corpus

Application of Artificial Intelligence Technology in Oncology: Towards the Establishment of Precision Medicine (2020, methodology): “domain adaptation methods such as gradient reversal layer [189], maximum mean discrepancy [190], and Generate To Adapt [191] for high-resolution medical images with complex feature distribution.”
Few-Shot Adversarial Domain Adaptation (2017, methodology): “The first two tasks includeM→ U , U →M, and followed the experimental setting in [12, 45, 33, 59, 49], which involves randomly selecting 2000 images from MNIST and 1800 images from USPS.”
Prototype-Guided Continual Adaptation for Class-Incremental Unsupervised Domain Adaptation (2022, methodology): “Step 5 [40,41,42,43,44,45,46,47,48,49] Pan, Screwdriver, Trash Can, Printer, Speaker, Eraser, Bucket, Chair, Calendar, Calculator”
Class-Incremental Domain Adaptation (2020, background): “The most popular UDA approach [11,25,30,46,49,52] is to learn a predictor h(x) = g ◦ f(x) having a domain-agnostic feature extractor f that is common to both the domains, and a classifier g which can be learned using source supervision.”
Unbiased Mean Teacher for Cross-domain Object Detection (2020, methodology): “Alternatively, adversarial training with domain classifier is also commonly used to learn domain-invariant representation [10, 11, 39, 48].”
Deep Subdomain Adaptation Network for Image Classification (2020, methodology): “For MNIST-USPS-SVHN, we compare DSAN with Domain Adversarial Neural Networks ( DANN ) [17], Deep Reconstruction-classiﬁcation Networks ( DRCN ) [51], Coupled Generative Adversarial Networks (CoGAN) [52], Adversarial Discriminative Domain Adaptation ( ADDA ) [30], Un-supervised Image-to-image Transla”
Unsupervised domain adaptation for medical imaging segmentation with self-ensembling (2018, methodology): “…domain-invariant feature spaces through auto-encoders (Ghifary et al., 2016), adversarial training (Ganin et al., 2016), GANs (Hoffman et al., 2017; Sankaranarayanan et al., 2018), or disentanglement strategies (Liu et al., 2018; Cao et al., 2018); (ii) methods based on the analysis of…”
Transferability vs. Discriminability: Batch Spectral Penalization for Adversarial Domain Adaptation (2019, background): “, 2017), Generate to Adapt (GTA) (Sankaranarayanan et al., 2018), Cycle-Consistent Adversarial Domain Adaptation (CyCADA) (Hoffman et al.”
MimicGAN: Robust Projection onto Image Manifolds with Corruption Mimicking (2019, neutral): “8 Gen to Adapt [49] 95.”
Under review as a conference paper at ICLR 2020 many domain adaptation methods (2019, methodology): “many domain adaptation methods Liu & Tuzel (2016); Sankaranarayanan et al. (2018); Volpi et al. (2018). CoGAN proposed by Liu & Tuzel (2016) learns the joint distribution of multiple domains without corresponding image pairs. Sankaranarayanan et al. (2018) propose the combined adversarial and discri”
Continuously Indexed Domain Adaptation (2020, background): “Much prior work has focused on the problem of domain adaptation (Zhao et al., 2017; Long et al., 2018; Saito et al., 2018; Sankaranarayanan et al., 2018; Zhang et al., 2019).”
Gradually Vanishing Bridge for Adversarial Domain Adaptation (2020, methodology): “Based on GAN, the transferable representations are constructed by the generator that tries to fool the discriminator in the minmax game [9, 17, 35, 36, 40].”
Detach and Adapt: Learning Cross-Domain Disentangled Deep Representation (2017, background): “GFK [8] JDA [22] SA [4] TJM [23] SCA [6] JGSA [33] DC [30] GR [5] CoGAN [20] ADDA [31] DRCN [7] ADGAN [28] CDRD”
Enhanced Transport Distance for Unsupervised Domain Adaptation (2020, methodology): “Sufﬁcient labeled data play an important role in training discriminative and interpretable models which have wide applications in pattern recognition and machine learning [24, 28].”
Minimum Class Confusion for Versatile Domain Adaptation (2019, neutral): “3 GTA [37] 89.”
Learning a Domain-Invariant Embedding for Unsupervised Domain Adaptation Using Class-Conditioned Distribution Alignment (2019, methodology): “against the recent adversarial learning algorithms: Generate to Adapt (GtA) [35], CoGAN [21], ADDA [41], CyCADA [17], and I2I-Adapt [24].”