The empirical fact that classifiers, trained on given data collections, perform poorly when tested on data acquired in different settings is theoretically explained in domain adaptation through a shift among distributions of the source and target domains. Alleviating the domain shift problem, especially in the challenging setting where no labeled data are available for the target domain, is paramount for having visual recognition systems working in the wild. As the problem stems from a shift among distributions, intuitively one should try to align them. In the literature, this has resulted in a stream of works attempting to align the feature representations learned from the source and target domains. Here we take a different route. Rather than introducing regularization terms aiming to promote the alignment of the two representations, we act at the distribution level through the introduction of \emphDomaIn Alignment Layers (\DIAL), able to match the observed source and target data distributions to a reference one. Thorough experiments on three different public benchmarks we confirm the power of our approach.
Just DIAL: DomaIn Alignment Layers for Unsupervised Domain Adaptation / Carlucci, FABIO MARIA; Porzi, Lorenzo; Caputo, Barbara; Ricci, Elisa; ROTA BULO', Samuel. - ELETTRONICO. - 1:(2017), pp. 87-100. (Intervento presentato al convegno International Conference on Image Analysis and Processing tenutosi a Catania, ITALY) [10.1007/978-3-319-68560-1_32].
Just DIAL: DomaIn Alignment Layers for Unsupervised Domain Adaptation
Fabio Maria Carlucci
;Lorenzo Porzi;Barbara Caputo;Elisa Ricci;Samuel Rota Bulò
2017
Abstract
The empirical fact that classifiers, trained on given data collections, perform poorly when tested on data acquired in different settings is theoretically explained in domain adaptation through a shift among distributions of the source and target domains. Alleviating the domain shift problem, especially in the challenging setting where no labeled data are available for the target domain, is paramount for having visual recognition systems working in the wild. As the problem stems from a shift among distributions, intuitively one should try to align them. In the literature, this has resulted in a stream of works attempting to align the feature representations learned from the source and target domains. Here we take a different route. Rather than introducing regularization terms aiming to promote the alignment of the two representations, we act at the distribution level through the introduction of \emphDomaIn Alignment Layers (\DIAL), able to match the observed source and target data distributions to a reference one. Thorough experiments on three different public benchmarks we confirm the power of our approach.File | Dimensione | Formato | |
---|---|---|---|
Carlucci_Just_2017.pdf
solo gestori archivio
Tipologia:
Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
571.65 kB
Formato
Adobe PDF
|
571.65 kB | Adobe PDF | Contatta l'autore |
Carlucci_copertina-frontespizio_Just_2017.pdf
solo gestori archivio
Tipologia:
Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
282.02 kB
Formato
Adobe PDF
|
282.02 kB | Adobe PDF | Contatta l'autore |
Carlucci_indice_Just_2017.pdf
solo gestori archivio
Tipologia:
Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
225.13 kB
Formato
Adobe PDF
|
225.13 kB | Adobe PDF | Contatta l'autore |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.