The design of neural network architectures is an important component for
achieving state-of-the-art performance with machine learning systems across a
broad array of tasks. Much work has endeavored to design and build
architectures automatically through clever construction of a search space
paired with simple learning algorithms. Recent progress has demonstrated that
such meta-learning methods may exceed scalable human-invented architectures on
image classification tasks. An open question is the degree to which such
methods may generalize to new domains. In this work we explore the construction
of meta-learning techniques for dense image prediction focused on the tasks of
scene parsing, person-part segmentation, and semantic image segmentation.
Constructing viable search spaces in this domain is challenging because of the
multi-scale representation of visual information and the necessity to operate
on high resolution imagery. Based on a survey of techniques in dense image
prediction, we construct a recursive search space and demonstrate that even
with efficient random search, we can identify architectures that outperform
human-invented architectures and achieve state-of-the-art performance on three
dense prediction tasks including 82.7\% on Cityscapes (street scene parsing),
71.3\% on PASCAL-Person-Part (person-part segmentation), and 87.9\% on PASCAL
VOC 2012 (semantic image segmentation). Additionally, the resulting
architecture is more computationally efficient, requiring half the parameters
and half the computational cost as previous state of the art systems.
%0 Generic
%1 chen2018searching
%A Chen, Liang-Chieh
%A Collins, Maxwell D.
%A Zhu, Yukun
%A Papandreou, George
%A Zoph, Barret
%A Schroff, Florian
%A Adam, Hartwig
%A Shlens, Jonathon
%D 2018
%K MultiScale final thema:neural_architecture_search
%T Searching for Efficient Multi-Scale Architectures for Dense Image
Prediction
%U http://arxiv.org/abs/1809.04184
%X The design of neural network architectures is an important component for
achieving state-of-the-art performance with machine learning systems across a
broad array of tasks. Much work has endeavored to design and build
architectures automatically through clever construction of a search space
paired with simple learning algorithms. Recent progress has demonstrated that
such meta-learning methods may exceed scalable human-invented architectures on
image classification tasks. An open question is the degree to which such
methods may generalize to new domains. In this work we explore the construction
of meta-learning techniques for dense image prediction focused on the tasks of
scene parsing, person-part segmentation, and semantic image segmentation.
Constructing viable search spaces in this domain is challenging because of the
multi-scale representation of visual information and the necessity to operate
on high resolution imagery. Based on a survey of techniques in dense image
prediction, we construct a recursive search space and demonstrate that even
with efficient random search, we can identify architectures that outperform
human-invented architectures and achieve state-of-the-art performance on three
dense prediction tasks including 82.7\% on Cityscapes (street scene parsing),
71.3\% on PASCAL-Person-Part (person-part segmentation), and 87.9\% on PASCAL
VOC 2012 (semantic image segmentation). Additionally, the resulting
architecture is more computationally efficient, requiring half the parameters
and half the computational cost as previous state of the art systems.
@misc{chen2018searching,
abstract = {The design of neural network architectures is an important component for
achieving state-of-the-art performance with machine learning systems across a
broad array of tasks. Much work has endeavored to design and build
architectures automatically through clever construction of a search space
paired with simple learning algorithms. Recent progress has demonstrated that
such meta-learning methods may exceed scalable human-invented architectures on
image classification tasks. An open question is the degree to which such
methods may generalize to new domains. In this work we explore the construction
of meta-learning techniques for dense image prediction focused on the tasks of
scene parsing, person-part segmentation, and semantic image segmentation.
Constructing viable search spaces in this domain is challenging because of the
multi-scale representation of visual information and the necessity to operate
on high resolution imagery. Based on a survey of techniques in dense image
prediction, we construct a recursive search space and demonstrate that even
with efficient random search, we can identify architectures that outperform
human-invented architectures and achieve state-of-the-art performance on three
dense prediction tasks including 82.7\% on Cityscapes (street scene parsing),
71.3\% on PASCAL-Person-Part (person-part segmentation), and 87.9\% on PASCAL
VOC 2012 (semantic image segmentation). Additionally, the resulting
architecture is more computationally efficient, requiring half the parameters
and half the computational cost as previous state of the art systems.},
added-at = {2020-12-19T19:30:29.000+0100},
author = {Chen, Liang-Chieh and Collins, Maxwell D. and Zhu, Yukun and Papandreou, George and Zoph, Barret and Schroff, Florian and Adam, Hartwig and Shlens, Jonathon},
biburl = {https://www.bibsonomy.org/bibtex/28fd178674e612b0311b2dd07a25b6b38/philipphaas},
description = {1809.04184.pdf},
interhash = {ff4b7f4359a03129b229124616d19ccb},
intrahash = {8fd178674e612b0311b2dd07a25b6b38},
keywords = {MultiScale final thema:neural_architecture_search},
note = {cite arxiv:1809.04184Comment: Accepted by NIPS 2018},
timestamp = {2021-01-20T12:46:36.000+0100},
title = {Searching for Efficient Multi-Scale Architectures for Dense Image
Prediction},
url = {http://arxiv.org/abs/1809.04184},
year = 2018
}