Zero-shot learning has gained popularity due to its potential to scale
recognition models without requiring additional training data. This is usually
achieved by associating categories with their semantic information like
attributes. However, we believe that the potential offered by this paradigm is
not yet fully exploited. In this work, we propose to utilize the structure of
the space spanned by the attributes using a set of relations. We devise
objective functions to preserve these relations in the embedding space, thereby
inducing semanticity to the embedding space. Through extensive experimental
evaluation on five benchmark datasets, we demonstrate that inducing semanticity
to the embedding space is beneficial for zero-shot learning. The proposed
approach outperforms the state-of-the-art on the standard zero-shot setting as
well as the more realistic generalized zero-shot setting. We also demonstrate
how the proposed approach can be useful for making approximate semantic
inferences about an image belonging to a category for which attribute
information is not available.
%0 Generic
%1 citeulike:14566375
%A xxx,
%D 2018
%K loss metric zero\_shot
%T Preserving Semantic Relations for Zero-Shot Learning
%U http://arxiv.org/abs/1803.03049
%X Zero-shot learning has gained popularity due to its potential to scale
recognition models without requiring additional training data. This is usually
achieved by associating categories with their semantic information like
attributes. However, we believe that the potential offered by this paradigm is
not yet fully exploited. In this work, we propose to utilize the structure of
the space spanned by the attributes using a set of relations. We devise
objective functions to preserve these relations in the embedding space, thereby
inducing semanticity to the embedding space. Through extensive experimental
evaluation on five benchmark datasets, we demonstrate that inducing semanticity
to the embedding space is beneficial for zero-shot learning. The proposed
approach outperforms the state-of-the-art on the standard zero-shot setting as
well as the more realistic generalized zero-shot setting. We also demonstrate
how the proposed approach can be useful for making approximate semantic
inferences about an image belonging to a category for which attribute
information is not available.
@misc{citeulike:14566375,
abstract = {{Zero-shot learning has gained popularity due to its potential to scale
recognition models without requiring additional training data. This is usually
achieved by associating categories with their semantic information like
attributes. However, we believe that the potential offered by this paradigm is
not yet fully exploited. In this work, we propose to utilize the structure of
the space spanned by the attributes using a set of relations. We devise
objective functions to preserve these relations in the embedding space, thereby
inducing semanticity to the embedding space. Through extensive experimental
evaluation on five benchmark datasets, we demonstrate that inducing semanticity
to the embedding space is beneficial for zero-shot learning. The proposed
approach outperforms the state-of-the-art on the standard zero-shot setting as
well as the more realistic generalized zero-shot setting. We also demonstrate
how the proposed approach can be useful for making approximate semantic
inferences about an image belonging to a category for which attribute
information is not available.}},
added-at = {2019-02-27T22:23:29.000+0100},
archiveprefix = {arXiv},
author = {xxx},
biburl = {https://www.bibsonomy.org/bibtex/21ca9603b6cff84ad0f996e39e4c3b511/nmatsuk},
citeulike-article-id = {14566375},
citeulike-linkout-0 = {http://arxiv.org/abs/1803.03049},
citeulike-linkout-1 = {http://arxiv.org/pdf/1803.03049},
day = 8,
eprint = {1803.03049},
interhash = {bc7fa024224a9f93988f01571c6d9d23},
intrahash = {1ca9603b6cff84ad0f996e39e4c3b511},
keywords = {loss metric zero\_shot},
month = mar,
posted-at = {2018-04-10 07:37:43},
priority = {0},
timestamp = {2019-02-27T22:23:29.000+0100},
title = {{Preserving Semantic Relations for Zero-Shot Learning}},
url = {http://arxiv.org/abs/1803.03049},
year = 2018
}