FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic
Segmentation
H. Wu, J. Zhang, K. Huang, K. Liang, and Y. Yu. (2019)cite arxiv:1903.11816Comment: Code is available in https://github.com/wuhuikai/FastFCN.
Abstract
Modern approaches for semantic segmentation usually employ dilated
convolutions in the backbone to extract high-resolution feature maps, which
brings heavy computation complexity and memory footprint. To replace the time
and memory consuming dilated convolutions, we propose a novel joint upsampling
module named Joint Pyramid Upsampling (JPU) by formulating the task of
extracting high-resolution feature maps into a joint upsampling problem. With
the proposed JPU, our method reduces the computation complexity by more than
three times without performance loss. Experiments show that JPU is superior to
other upsampling modules, which can be plugged into many existing approaches to
reduce computation complexity and improve performance. By replacing dilated
convolutions with the proposed JPU module, our method achieves the
state-of-the-art performance in Pascal Context dataset (mIoU of 53.13%) and
ADE20K dataset (final score of 0.5584) while running 3 times faster.
Description
[1903.11816] FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation
%0 Generic
%1 wu2019fastfcn
%A Wu, Huikai
%A Zhang, Junge
%A Huang, Kaiqi
%A Liang, Kongming
%A Yu, Yizhou
%D 2019
%K semanticsegmentation
%T FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic
Segmentation
%U http://dblp.uni-trier.de/db/journals/corr/corr1903.html#abs-1903-11816
%X Modern approaches for semantic segmentation usually employ dilated
convolutions in the backbone to extract high-resolution feature maps, which
brings heavy computation complexity and memory footprint. To replace the time
and memory consuming dilated convolutions, we propose a novel joint upsampling
module named Joint Pyramid Upsampling (JPU) by formulating the task of
extracting high-resolution feature maps into a joint upsampling problem. With
the proposed JPU, our method reduces the computation complexity by more than
three times without performance loss. Experiments show that JPU is superior to
other upsampling modules, which can be plugged into many existing approaches to
reduce computation complexity and improve performance. By replacing dilated
convolutions with the proposed JPU module, our method achieves the
state-of-the-art performance in Pascal Context dataset (mIoU of 53.13%) and
ADE20K dataset (final score of 0.5584) while running 3 times faster.
@misc{wu2019fastfcn,
abstract = {Modern approaches for semantic segmentation usually employ dilated
convolutions in the backbone to extract high-resolution feature maps, which
brings heavy computation complexity and memory footprint. To replace the time
and memory consuming dilated convolutions, we propose a novel joint upsampling
module named Joint Pyramid Upsampling (JPU) by formulating the task of
extracting high-resolution feature maps into a joint upsampling problem. With
the proposed JPU, our method reduces the computation complexity by more than
three times without performance loss. Experiments show that JPU is superior to
other upsampling modules, which can be plugged into many existing approaches to
reduce computation complexity and improve performance. By replacing dilated
convolutions with the proposed JPU module, our method achieves the
state-of-the-art performance in Pascal Context dataset (mIoU of 53.13%) and
ADE20K dataset (final score of 0.5584) while running 3 times faster.},
added-at = {2020-03-18T16:57:59.000+0100},
author = {Wu, Huikai and Zhang, Junge and Huang, Kaiqi and Liang, Kongming and Yu, Yizhou},
biburl = {https://www.bibsonomy.org/bibtex/2df3dc3a5d5525b942125b0d9b3bef28a/annakrause},
description = {[1903.11816] FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation},
interhash = {86bfc58ea5c86d04db1e07fc70afef6d},
intrahash = {df3dc3a5d5525b942125b0d9b3bef28a},
keywords = {semanticsegmentation},
note = {cite arxiv:1903.11816Comment: Code is available in https://github.com/wuhuikai/FastFCN},
timestamp = {2020-03-27T09:10:59.000+0100},
title = {FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic
Segmentation},
url = {http://dblp.uni-trier.de/db/journals/corr/corr1903.html#abs-1903-11816},
year = 2019
}