Structure Preserving Generative Cross-Domain Learning

Xia, Haifeng; Ding, Zhengming

doi:10.1109/cvpr42600.2020.00442

Cited by 37 publications

(12 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Specifically, TSA empowers BSP with 4.9% improvement, achieving the highest accuracy 71.2% on average. Based on these promising re- [20] 92.1 100.0 100.0 88.0 71.0 67.8 86.2 MCD [41] 88.6 98.5 100.0 92.2 69.5 69.7 86.5 BNM [7] 91.5 98.5 100.0 90.3 70.9 71.6 87.1 DMRL [48] 90.8 99.0 100.0 93.4 73.0 71.2 87.9 SymNets [54] 90.8 98.8 100.0 93.9 74.6 72.5 88.4 TAT [23] 92.5 99.3 100.0 93.2 73.1 72.1 88.4 MDD [53] 94.5 98.4 100.0 93.5 74.6 72.2 88.9 GVB-GD [8] 94.8 98.7 100.0 95.0 73.4 73.7 89.3 GSP [49] 92.9 98.7 99.8 94.5 75.9 74.9 89.5 ResNet-50 [13] sults, we can infer that TSA can stably enhance the transferability of classifiers on this difficult cross-domain dataset.…”

Section: Resultsmentioning

confidence: 99%

Transferable Semantic Augmentation for Domain Adaptation

Li¹,

Xie²,

Gong³

et al. 2021

Preprint

View full text Add to dashboard Cite

Domain adaptation has been widely explored by transferring the knowledge from a label-rich source domain to a related but unlabeled target domain. Most existing domain adaptation algorithms attend to adapting feature representations across two domains with the guidance of a shared source-supervised classifier. However, such classifier limits the generalization ability towards unlabeled target recognition. To remedy this, we propose a Transferable Semantic Augmentation (TSA) approach to enhance the classifier adaptation ability through implicitly generating source features towards target semantics. Specifically, TSA is inspired by the fact that deep feature transformation towards a certain direction can be represented as meaningful semantic altering in the original input space. Thus, source features can be augmented to effectively equip with target semantics to train a more transferable classifier. To achieve this, for each class, we first use the inter-domain feature mean difference and target intra-class feature covariance to construct a multivariate normal distribution. Then we augment source features with random directions sampled from the distribution class-wisely. Interestingly, such source augmentation is implicitly implemented through an expected transferable cross-entropy loss over the augmented source distribution, where an upper bound of the expected loss is derived and minimized, introducing negligible computational overhead. As a light-weight and general technique, TSA can be easily plugged into various domain adaptation methods, bringing remarkable improvements. Comprehensive experiments on cross-domain benchmarks validate the efficacy of TSA.

show abstract

Section: Resultsmentioning

confidence: 99%

Transferable Semantic Augmentation for Domain Adaptation

Li¹,

Xie²,

Gong³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…To this end, [24] proposes to translate the source data to target data for domain adaptation in segmentation task. Several researches [25]- [32] focus on utilizing generative adversarial architecture to transfer domaininvariant knowledge across domains in the feature space. Besides, Tsai et al [33] develop a multi-level contextual adaptation framework in the output space.…”

Section: B Semantic Knowledge Transfermentioning

confidence: 99%

Weakly-Supervised Cross-Domain Adaptation for Endoscopic Lesions Segmentation

Dong

Cong

Sun

et al. 2021

IEEE Trans. Circuits Syst. Video Technol.

Self Cite

View full text Add to dashboard Cite

Weakly-supervised learning has attracted growing research attention on medical lesions segmentation due to significant saving in pixel-level annotation cost. However, 1) most existing methods require effective prior and constraints to explore the intrinsic lesions characterization, which only generates incorrect and rough prediction; 2) they neglect the underlying semantic dependencies among weakly-labeled target enteroscopy diseases and fully-annotated source gastroscope lesions, while forcefully utilizing untransferable dependencies leads to the negative performance. To tackle above issues, we propose a new weakly-supervised lesions transfer framework, which can not only explore transferable domain-invariant knowledge across different datasets, but also prevent the negative transfer of untransferable representations. Specifically, a Wasserstein quantified transferability framework is developed to highlight widerange transferable contextual dependencies, while neglecting the irrelevant semantic characterizations. Moreover, a novel selfsupervised pseudo label generator is designed to equally provide confident pseudo pixel labels for both hard-to-transfer and easyto-transfer target samples. It inhibits the enormous deviation of false pseudo pixel labels under the self-supervision manner. Afterwards, dynamically-searched feature centroids are aligned to narrow category-wise distribution shift. Comprehensive theoretical analysis and experiments show the superiority of our model on the endoscopic dataset and several public datasets.

show abstract

“…Since the target domain is unlabeled, these works rely on predicting pseudo-labeling (Kang et al, 2019) or computing prototype representations of source and target classes (Wang & Breckon, 2020), and then the target domain samples are classified by the prototype of the target domain classes during the training process. Structure-preserving methods (Ren et al, 2019;Xia & Ding, 2020) try to achieve the class-level transfer by matching the structure graphs across domains. However, as observed by Chen et al (2019), the discrim-inability may be decreased when the models only focus on enhancing transferability.…”

Section: Class-specific Learningmentioning

confidence: 99%

Discriminative and Geometry-Aware Unsupervised Domain Adaptation

Luo

Chen

et al. 2020

IEEE Trans. Cybern.

View full text Add to dashboard Cite

Domain adaptation (DA) aims to generalize a learning model across training and testing data despite the mismatch of their data distributions. In light of a theoretical estimation of upper error bound, we argue in this paper that an effective DA method should 1) search a shared feature subspace where source and target data are not only aligned in terms of distributions as most state of the art DA methods do, but also discriminative in that instances of different classes are well separated; 2) account for the geometric structure of the underlying data manifold when inferring data labels on the target domain. In comparison with a baseline DA method which only cares about data distribution alignment between source and target, we derive three different DA models, namely CDDA, GA-DA, and DGA-DA, to highlight the contribution of Close yet Discriminative DA(CDDA) based on 1), Geometry Aware DA (GA-DA) based on 2), and finally Discriminative and Geometry Aware DA (DGA-DA) implementing jointly 1) and 2). Using both synthetic and real data, we show the effectiveness of the proposed approach which consistently outperforms state of the art DA methods over 36 image classification DA tasks through 6 popular benchmarks. We further carry out in-depth analysis of the proposed DA method in quantifying the contribution of each term of our DA model and provide insights into the proposed DA methods in visualizing both real and synthetic data.

show abstract

Structure Preserving Generative Cross-Domain Learning

Cited by 37 publications

References 28 publications

Transferable Semantic Augmentation for Domain Adaptation

Transferable Semantic Augmentation for Domain Adaptation

Weakly-Supervised Cross-Domain Adaptation for Endoscopic Lesions Segmentation

Discriminative and Geometry-Aware Unsupervised Domain Adaptation

Contact Info

Product

Resources

About