Filters








103,162 Hits in 3.7 sec

Learning a Domain-Invariant Embedding for Unsupervised Domain Adaptation Using Class-Conditioned Distribution Alignment [article]

Alex Gabourie, Mohammad Rostami, Philip Pope, Soheil Kolouri, Kyungnam Kim
2019 arXiv   pre-print
To mitigate the challenge of class matching, we also align corresponding classes in the embedding space by using high confidence pseudo-labels for the target domain, i.e. assigning the class for which  ...  We address the problem of unsupervised domain adaptation (UDA) by learning a cross-domain agnostic embedding space, where the distance between the probability distributions of the two source and target  ...  A second group of domain adaptation algorithms match the distributions directly in the embedding by using a shared cross-domain mapping such that the distance between the two distributions is minimized  ... 
arXiv:1907.02271v2 fatcat:2ct3ilhgenetlmtqncjylk5nxm

Beyond the Deep Metric Learning: Enhance the Cross-Modal Matching with Adversarial Discriminative Domain Regularization [article]

Li Ren, Kai Li, LiQiang Wang, Kien Hua
2020 arXiv   pre-print
Matching information across image and text modalities is a fundamental challenge for many applications that involve both vision and natural language processing.  ...  Existing approaches mainly match the local visual objects and the sentence words in a shared space with attention mechanisms.  ...  Note that we denote the embedding networks (or the feature generators) for both domains as g θ (·) for convenience.  ... 
arXiv:2010.12126v2 fatcat:we74xd3jdzdzlev2fewj7spf7m

Unsupervised Domain Adaptation with Imbalanced Cross-Domain Data

Tzu Ming Harry Hsu, Wei Yu Chen, Cheng-An Hou, Yao-Hung Hubert Tsai, Yi-Ren Yeh, Yu-Chiang Frank Wang
2015 2015 IEEE International Conference on Computer Vision (ICCV)  
For standard unsupervised domain adaptation, one typically obtains labeled data in the source domain and only observes unlabeled data in the target domain.  ...  We address a challenging unsupervised domain adaptation problem with imbalanced cross-domain data.  ...  data embedding for the adapted data using t-distributed stochastic neighbor embedding (t-SNE) [26] .  ... 
doi:10.1109/iccv.2015.469 dblp:conf/iccv/HsuCHTYW15 fatcat:peyg4gxotzc7tpowtakoxgvhmu

Domain Adaptation via Bregman divergence minimization

Mozhdeh Zandifar, Shiva Noori Saray, Jafar Tahmoresnezhad
2021 Scientia Iranica. International Journal of Science and Technology  
DAB is designed based on the constraints of FLDA, with the aim of the coupled marginal and conditional distribution shifts adaptation through Bregman divergence minimization.  ...  However, when the learning data (source domain) have a different distribution compared with the testing data (target domain), the FLDA-based models may not work well, and the performance degrades, dramatically  ...  domain adaptation via Bregman divergence minimization (DAB), visual domain adaptation (VDA), transfer joint matching (TJM) and joint distribution adaptation (JDA).  ... 
doi:10.24200/sci.2021.51486.2210 fatcat:xlwuk7kikffirkr5t6yesn7ylq

Adaptive RGB Image Recognition by Visual-Depth Embedding

Ziyun Cai, Yang Long, Ling Shao
2018 IEEE Transactions on Image Processing  
At last, aVDE models two separate learning strategies for domain adaptation (feature matching and instance reweighting) in a unified optimization problem, which matches features and reweights instances  ...  jointly across the shared latent space and the projected target domain for an adaptive classifier.  ...  Both of them focus on RGB source domain and RGB target domain, which is a completely different task. ii) Our NMFrelated equations are designed for visual-depth embedding, not for domain adaptation. iii  ... 
doi:10.1109/tip.2018.2806839 pmid:29994784 fatcat:7ie5bp6jsrhyvistkv6ftixeyi

SIGMA: Semantic-complete Graph Matching for Domain Adaptive Object Detection [article]

Wuyang Li, Xinyu Liu, Yixuan Yuan
2022 arXiv   pre-print
To overcome these challenges, we propose a novel SemantIc-complete Graph MAtching (SIGMA) framework for DAOD, which completes mismatched semantics and reformulates the adaptation with graph matching.  ...  Domain Adaptive Object Detection (DAOD) leverages a labeled domain to learn an object detector generalizing to a novel domain free of annotations.  ...  also be aligned for domain adaptation.  ... 
arXiv:2203.06398v3 fatcat:tmx7fu4wnvghnn2tb7mj2ij6sa

Self-Rule to Multi-Adapt: Generalized Multi-source Feature Learning Using Unsupervised Domain Adaptation for Colorectal Cancer Tissue Detection [article]

Christian Abbet, Linda Studer, Andreas Fischer, Heather Dawson, Inti Zlobec, Behzad Bozorgtabar, Jean-Philippe Thiran
2022 arXiv   pre-print
Making use of open-source data for pre-training or using domain adaptation can be a way to overcome this issue.  ...  Our method harnesses both domains' structures by capturing visual similarity with intra-domain and cross-domain self-supervision.  ...  Philipp Zens, and Guillaume Vray for the annotation of the WSI crops, the feedback on complex stroma detection and the computation of a few baselines that greatly helped the evaluation of our method.  ... 
arXiv:2108.09178v2 fatcat:nhoqz2gx4rbwlbcjzcfcpfevga

Sequential Model Adaptation Using Domain Agnostic Internal Distributions [article]

Mohammad Rostami, Aram Galstyan
2021 arXiv   pre-print
We develop an algorithm for sequential adaptation of a classifier that is trained for a source domain to generalize in an unannotated target domain.  ...  We align the distributions of the source and the target domains in a discriminative embedding space via an intermediate internal distribution.  ...  In order to adapt the model to work well for the target domain, we update the model such that the encoder matches the target distribution into the internal distribution in the embedding space.  ... 
arXiv:2007.00197v4 fatcat:qlv4zgkoira2jglhcxvkwwpn74

Deep Transfer Learning for Few-Shot SAR Image Classification

Mohammad Rostami, Soheil Kolouri, Eric Eaton, Kyungnam Kim
2019 Remote Sensing  
We transfer knowledge from the EO domain through learning a shared invariant cross-domain embedding space that is also discriminative for classification.  ...  We use the Sliced Wasserstein Distance (SWD) to measure and minimize the distance between these two distributions and use a limited number of SAR label data points to match the distributions class-conditionally  ...  We have presented the Umap visualization of the datasets in the embedding space for a particular experiment in Figure 5 .  ... 
doi:10.3390/rs11111374 fatcat:xngpe5eporevzdyud26w6iurpy

Unsupervised Domain Adaptation for Zero-Shot Learning

Elyor Kodirov, Tao Xiang, Zhenyong Fu, Shaogang Gong
2015 2015 IEEE International Conference on Computer Vision (ICCV)  
A ZSL method typically assumes that the two domains share a common semantic representation space, where a visual feature vector extracted from an image/video can be projected/embedded using a projection  ...  guidance for the knowledge transfer.  ...  Our regularised sparse coding based domain adaptation framework combines the visual feature projection and visual-semantic similarity matching based approaches in a single formulation.  ... 
doi:10.1109/iccv.2015.282 dblp:conf/iccv/KodirovXFG15 fatcat:r2o6prdiabbileemh4dmtddcoq

Transferrable Prototypical Networks for Unsupervised Domain Adaptation

Yingwei Pan, Ting Yao, Yehao Li, Yu Wang, Chong-Wah Ngo, Tao Mei
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Specifically, we present Transferrable Prototypical Networks (TPN) for adaptation such that the prototypes for each class in source and target domains are close in the embedding space and the score distributions  ...  In this paper, we introduce a new idea for unsupervised domain adaptation via a remold of Prototypical Networks, which learn an embedding space and perform classification via a remold of the distances  ...  In this way, the general-purpose adaptation is to represent each class distribution by a prototype and match the prototypes of each class in the embedding space learnt on the data from different domains  ... 
doi:10.1109/cvpr.2019.00234 dblp:conf/cvpr/PanYLWNM19 fatcat:uijemqaydjcf5e4t2tnbybnhje

Transferrable Prototypical Networks for Unsupervised Domain Adaptation [article]

Yingwei Pan, Ting Yao, Yehao Li, Yu Wang, Chong-Wah Ngo, Tao Mei
2019 arXiv   pre-print
Specifically, we present Transferrable Prototypical Networks (TPN) for adaptation such that the prototypes for each class in source and target domains are close in the embedding space and the score distributions  ...  In this paper, we introduce a new idea for unsupervised domain adaptation via a remold of Prototypical Networks, which learn an embedding space and perform classification via a remold of the distances  ...  In this way, the general-purpose adaptation is to represent each class distribution by a prototype and match the prototypes of each class in the embedding space learnt on the data from different domains  ... 
arXiv:1904.11227v1 fatcat:eveiyypusbenhjkogdie53ubem

Cross-Dataset Adaptation for Visual Question Answering

Fei Sha, Hexiang Hu, Wei-Lun Chao
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
Analogous to domain adaptation for visual recognition, this setting is appealing when the target dataset does not have a sufficient amount of labeled data to learn an "in-domain" model.  ...  We investigate the problem of cross-dataset adaptation for visual question answering (Visual QA). Our goal is to train a Visual QA model on a source dataset but apply it to another target one.  ...  In Sect. 4, we define tasks of domain adaptation for Visual QA. In Sect. 4.2, we describe the proposed domain adaptation algorithm.  ... 
doi:10.1109/cvpr.2018.00599 dblp:conf/cvpr/ChaoHS18 fatcat:dleg5tpld5f3xdvhyejvu6bb2a

Mind the Gap: Domain Gap Control for Single Shot Domain Adaptation for Generative Adversarial Networks [article]

Peihao Zhu, Rameen Abdal, John Femiani, Peter Wonka
2021 arXiv   pre-print
We present a new method for one shot domain adaptation. The input to our method is trained GAN that can produce images in domain A and a single reference image I_B from domain B.  ...  Second, our solution allows for more degrees of freedom to control the domain gap, i.e. what aspects of image I_B are used to define the domain B.  ...  We show additional visual results in the appendix, including results on cars and dogs and results for fine-tuning the domain adaptation. User Study.  ... 
arXiv:2110.08398v2 fatcat:ilfin2ybjrgxlhz3wblmyti6jq

Multi-Task Zero-Shot Action Recognition with Prioritised Data Augmentation [chapter]

Xun Xu, Timothy M. Hospedales, Shaogang Gong
2016 Lecture Notes in Computer Science  
Zero-Shot Learning (ZSL) promises to scale visual recognition by bypassing the conventional model training requirement of annotated examples for every category.  ...  However, existing ZSL methods suffer from auxiliary-target domain shift intrinsically induced by assuming the same mapping for the disjoint auxiliary and target classes.  ...  Distributed Space Matching Given a trained visual-semantic regression f , we project testing set visual feature x te into the semantic label embedding space.  ... 
doi:10.1007/978-3-319-46475-6_22 fatcat:zmwhktds3vfndj6wh364qpsutu
« Previous Showing results 1 — 15 out of 103,162 results