Matches in SemOpenAlex for { <https://semopenalex.org/work/W3201612248> ?p ?o ?g. }
- W3201612248 abstract "Unsupervised domain adaptation (UDA) aims to transfer knowledge learned from a labeled source domain to a different unlabeled target domain. Most existing UDA methods focus on learning domain-invariant feature representation, either from the domain level or category level, using convolution neural networks (CNNs)-based frameworks. One fundamental problem for the category level based UDA is the production of pseudo labels for samples in target domain, which are usually too noisy for accurate domain alignment, inevitably compromising the UDA performance. With the success of Transformer in various tasks, we find that the cross-attention in Transformer is robust to the noisy input pairs for better feature alignment, thus in this paper Transformer is adopted for the challenging UDA task. Specifically, to generate accurate input pairs, we design a two-way center-aware labeling algorithm to produce pseudo labels for target samples. Along with the pseudo labels, a weight-sharing triple-branch transformer framework is proposed to apply self-attention and cross-attention for source/target feature learning and source-target domain alignment, respectively. Such design explicitly enforces the framework to learn discriminative domain-specific and domain-invariant representations simultaneously. The proposed method is dubbed CDTrans (cross-domain transformer), and it provides one of the first attempts to solve UDA tasks with a pure transformer solution. Experiments show that our proposed method achieves the best performance on public UDA datasets, e.g. VisDA-2017 and DomainNet. Code and models are available at https://github.com/CDTrans/CDTrans." @default.
- W3201612248 created "2021-09-27" @default.
- W3201612248 creator A5013097653 @default.
- W3201612248 creator A5019560977 @default.
- W3201612248 creator A5020103264 @default.
- W3201612248 creator A5038836690 @default.
- W3201612248 creator A5042680345 @default.
- W3201612248 creator A5069394608 @default.
- W3201612248 date "2021-09-13" @default.
- W3201612248 modified "2023-09-23" @default.
- W3201612248 title "CDTrans: Cross-domain Transformer for Unsupervised Domain Adaptation" @default.
- W3201612248 cites W1565327149 @default.
- W3201612248 cites W1722318740 @default.
- W3201612248 cites W1821462560 @default.
- W3201612248 cites W2125865219 @default.
- W3201612248 cites W2159291411 @default.
- W3201612248 cites W2194775991 @default.
- W3201612248 cites W2293363371 @default.
- W3201612248 cites W2478454054 @default.
- W3201612248 cites W2584009249 @default.
- W3201612248 cites W2590953969 @default.
- W3201612248 cites W2593768305 @default.
- W3201612248 cites W2627183927 @default.
- W3201612248 cites W2766897166 @default.
- W3201612248 cites W2767382337 @default.
- W3201612248 cites W2794887779 @default.
- W3201612248 cites W2795155917 @default.
- W3201612248 cites W2798681837 @default.
- W3201612248 cites W2883725317 @default.
- W3201612248 cites W2895281799 @default.
- W3201612248 cites W2903739847 @default.
- W3201612248 cites W2946812986 @default.
- W3201612248 cites W2948959975 @default.
- W3201612248 cites W2949813473 @default.
- W3201612248 cites W2962687275 @default.
- W3201612248 cites W2962808524 @default.
- W3201612248 cites W2963275094 @default.
- W3201612248 cites W2963341956 @default.
- W3201612248 cites W2963393201 @default.
- W3201612248 cites W2963403868 @default.
- W3201612248 cites W2963532621 @default.
- W3201612248 cites W2963826681 @default.
- W3201612248 cites W2964051877 @default.
- W3201612248 cites W2964278684 @default.
- W3201612248 cites W2964288524 @default.
- W3201612248 cites W2966743431 @default.
- W3201612248 cites W2970092410 @default.
- W3201612248 cites W2970987681 @default.
- W3201612248 cites W2980096013 @default.
- W3201612248 cites W2981720610 @default.
- W3201612248 cites W2982204955 @default.
- W3201612248 cites W2991405316 @default.
- W3201612248 cites W2997739323 @default.
- W3201612248 cites W3025988480 @default.
- W3201612248 cites W3034401437 @default.
- W3201612248 cites W3034526587 @default.
- W3201612248 cites W3035576098 @default.
- W3201612248 cites W3039883906 @default.
- W3201612248 cites W3094277917 @default.
- W3201612248 cites W3094502228 @default.
- W3201612248 cites W3102758664 @default.
- W3201612248 cites W3109093849 @default.
- W3201612248 cites W3118589616 @default.
- W3201612248 cites W3118802333 @default.
- W3201612248 cites W3119997354 @default.
- W3201612248 cites W3121523901 @default.
- W3201612248 cites W3128723389 @default.
- W3201612248 cites W3131500599 @default.
- W3201612248 cites W3133696297 @default.
- W3201612248 cites W3136635488 @default.
- W3201612248 cites W3138516171 @default.
- W3201612248 cites W3139633126 @default.
- W3201612248 cites W3140576409 @default.
- W3201612248 cites W3143320354 @default.
- W3201612248 cites W3151130473 @default.
- W3201612248 cites W3164024107 @default.
- W3201612248 cites W3165550458 @default.
- W3201612248 cites W3168101492 @default.
- W3201612248 cites W3168124404 @default.
- W3201612248 cites W3169909366 @default.
- W3201612248 cites W3169938586 @default.
- W3201612248 cites W3170874841 @default.
- W3201612248 cites W3171206729 @default.
- W3201612248 cites W3173631098 @default.
- W3201612248 cites W3175370657 @default.
- W3201612248 cites W3175452902 @default.
- W3201612248 cites W3176720610 @default.
- W3201612248 cites W3177117566 @default.
- W3201612248 cites W3177183540 @default.
- W3201612248 cites W3181127262 @default.
- W3201612248 cites W3187418919 @default.
- W3201612248 cites W3188427387 @default.
- W3201612248 cites W3190216403 @default.
- W3201612248 cites W3192174868 @default.
- W3201612248 cites W3195000282 @default.
- W3201612248 doi "https://doi.org/10.48550/arxiv.2109.06165" @default.
- W3201612248 hasPublicationYear "2021" @default.
- W3201612248 type Work @default.
- W3201612248 sameAs 3201612248 @default.
- W3201612248 citedByCount "2" @default.