Natural images contain many variations such as illumination differences,
affine transformations, and shape distortions. Correctly classifying these
variations poses a long standing problem. The most commonly adopted solution is
to build large-scale datasets that contain objects under different variations.
However, this approach is not ideal since it is computationally expensive and
it is hard to cover all variations in one single dataset. Towards addressing
this difficulty, we propose the spatial transformer introspective neural
network (ST-INN) that explicitly generates samples with the unseen affine
transformation variations in the training set. Experimental results indicate
ST-INN achieves classification accuracy improvements on several benchmark
datasets, including MNIST, affNIST, SVHN and CIFAR-10. We further extend our
method to cross dataset classification tasks and few-shot learning problems to
verify our method under extreme conditions and observe substantial improvements
from experiment results.
%0 Generic
%1 citeulike:14608951
%A xxx,
%D 2018
%K attention augmentation classification loss
%T Spatial Transformer Introspective Neural Network
%U http://arxiv.org/abs/1805.06447
%X Natural images contain many variations such as illumination differences,
affine transformations, and shape distortions. Correctly classifying these
variations poses a long standing problem. The most commonly adopted solution is
to build large-scale datasets that contain objects under different variations.
However, this approach is not ideal since it is computationally expensive and
it is hard to cover all variations in one single dataset. Towards addressing
this difficulty, we propose the spatial transformer introspective neural
network (ST-INN) that explicitly generates samples with the unseen affine
transformation variations in the training set. Experimental results indicate
ST-INN achieves classification accuracy improvements on several benchmark
datasets, including MNIST, affNIST, SVHN and CIFAR-10. We further extend our
method to cross dataset classification tasks and few-shot learning problems to
verify our method under extreme conditions and observe substantial improvements
from experiment results.
@misc{citeulike:14608951,
abstract = {{Natural images contain many variations such as illumination differences,
affine transformations, and shape distortions. Correctly classifying these
variations poses a long standing problem. The most commonly adopted solution is
to build large-scale datasets that contain objects under different variations.
However, this approach is not ideal since it is computationally expensive and
it is hard to cover all variations in one single dataset. Towards addressing
this difficulty, we propose the spatial transformer introspective neural
network (ST-INN) that explicitly generates samples with the unseen affine
transformation variations in the training set. Experimental results indicate
ST-INN achieves classification accuracy improvements on several benchmark
datasets, including MNIST, affNIST, SVHN and CIFAR-10. We further extend our
method to cross dataset classification tasks and few-shot learning problems to
verify our method under extreme conditions and observe substantial improvements
from experiment results.}},
added-at = {2019-02-27T22:23:29.000+0100},
archiveprefix = {arXiv},
author = {xxx},
biburl = {https://www.bibsonomy.org/bibtex/28664d1d6f1f1ee8d6265c7f9179728f6/nmatsuk},
citeulike-article-id = {14608951},
citeulike-linkout-0 = {http://arxiv.org/abs/1805.06447},
citeulike-linkout-1 = {http://arxiv.org/pdf/1805.06447},
day = 16,
eprint = {1805.06447},
interhash = {d41aa9c9f1a7a7175a882d823c2a10d3},
intrahash = {8664d1d6f1f1ee8d6265c7f9179728f6},
keywords = {attention augmentation classification loss},
month = may,
posted-at = {2018-06-28 08:34:13},
priority = {2},
timestamp = {2019-02-27T22:23:29.000+0100},
title = {{Spatial Transformer Introspective Neural Network}},
url = {http://arxiv.org/abs/1805.06447},
year = 2018
}