Gated Path Selection Network for Semantic Segmentation

Qichuan Geng; Hong Zhang; Xiaojuan Qi; Gao Huang; Ruigang Yang; Zhong Zhou

doi:10.1109/TIP.2020.3046921

Gated Path Selection Network for Semantic Segmentation

Qichuan Geng, Hong Zhang, Xiaojuan Qi, Gao Huang, Ruigang Yang, Zhong Zhou

Computer Science

Research output: Contribution to journal › Article › peer-review

26 Scopus citations

Abstract

Semantic segmentation is a challenging task that needs to handle large scale variations, deformations, and different viewpoints. In this paper, we develop a novel network named Gated Path Selection Network (GPSNet), which aims to adaptively select receptive fields while maintaining the dense sampling capability. In GPSNet, we first design a two-dimensional SuperNet, which densely incorporates features from growing receptive fields. And then, a Comparative Feature Aggregation (CFA) module is introduced to dynamically aggregate discriminative semantic context. In contrast to previous works that focus on optimizing sparse sampling locations on regular grids, GPSNet can adaptively harvest free form dense semantic context information. The derived adaptive receptive fields and dense sampling locations are data-dependent and flexible which can model various contexts of objects. On two representative semantic segmentation datasets, i.e., Cityscapes and ADE20K, we show that the proposed approach consistently outperforms previous methods without bells and whistles.

Original language	English
Article number	9318517
Pages (from-to)	2436-2449
Number of pages	14
Journal	IEEE Transactions on Image Processing
Volume	30
DOIs	https://doi.org/10.1109/TIP.2020.3046921
State	Published - 2021

Bibliographical note

Publisher Copyright:
© 2021 IEEE.

Funding

Manuscript received April 2, 2020; revised October 10, 2020; accepted December 5, 2020. Date of publication January 8, 2021; date of current version February 1, 2021. This work was supported in part by the National Key Research and Development Program of China under Grant 2018YFB2100601 and in part by the National Natural Science Foundation of China (NSFC) under Grant 61872024. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Raja Bala. (Corresponding author: Zhong Zhou.) Qichuan Geng is with the State Key Laboratory of Virtual Reality Technology and Systems, School of Computer Science and Engineering, Beihang University, Beijing 100191, China. His work was partially done when he was an intern at Baidu Research, Beijing 100193, China (e-mail: zhaokefirst@ buaa.edu.cn).

Funders	Funder number
National Natural Science Foundation of China (NSFC)	61872024
National Key Research and Development Program of China	2018YFB2100601

Keywords

Semantic segmentation
adaptive context aggregation
adaptive receptive fields and sampling locations
local discriminative feature

ASJC Scopus subject areas

Software
Computer Graphics and Computer-Aided Design

Access to Document

10.1109/TIP.2020.3046921

Cite this

@article{46599c61d38849cb9f45c3e71066fd47,

title = "Gated Path Selection Network for Semantic Segmentation",

abstract = "Semantic segmentation is a challenging task that needs to handle large scale variations, deformations, and different viewpoints. In this paper, we develop a novel network named Gated Path Selection Network (GPSNet), which aims to adaptively select receptive fields while maintaining the dense sampling capability. In GPSNet, we first design a two-dimensional SuperNet, which densely incorporates features from growing receptive fields. And then, a Comparative Feature Aggregation (CFA) module is introduced to dynamically aggregate discriminative semantic context. In contrast to previous works that focus on optimizing sparse sampling locations on regular grids, GPSNet can adaptively harvest free form dense semantic context information. The derived adaptive receptive fields and dense sampling locations are data-dependent and flexible which can model various contexts of objects. On two representative semantic segmentation datasets, i.e., Cityscapes and ADE20K, we show that the proposed approach consistently outperforms previous methods without bells and whistles.",

keywords = "Semantic segmentation, adaptive context aggregation, adaptive receptive fields and sampling locations, local discriminative feature",

author = "Qichuan Geng and Hong Zhang and Xiaojuan Qi and Gao Huang and Ruigang Yang and Zhong Zhou",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.",

year = "2021",

doi = "10.1109/TIP.2020.3046921",

language = "English",

volume = "30",

pages = "2436--2449",

}

TY - JOUR

T1 - Gated Path Selection Network for Semantic Segmentation

AU - Geng, Qichuan

AU - Zhang, Hong

AU - Qi, Xiaojuan

AU - Huang, Gao

AU - Yang, Ruigang

AU - Zhou, Zhong

PY - 2021

Y1 - 2021

N2 - Semantic segmentation is a challenging task that needs to handle large scale variations, deformations, and different viewpoints. In this paper, we develop a novel network named Gated Path Selection Network (GPSNet), which aims to adaptively select receptive fields while maintaining the dense sampling capability. In GPSNet, we first design a two-dimensional SuperNet, which densely incorporates features from growing receptive fields. And then, a Comparative Feature Aggregation (CFA) module is introduced to dynamically aggregate discriminative semantic context. In contrast to previous works that focus on optimizing sparse sampling locations on regular grids, GPSNet can adaptively harvest free form dense semantic context information. The derived adaptive receptive fields and dense sampling locations are data-dependent and flexible which can model various contexts of objects. On two representative semantic segmentation datasets, i.e., Cityscapes and ADE20K, we show that the proposed approach consistently outperforms previous methods without bells and whistles.

AB - Semantic segmentation is a challenging task that needs to handle large scale variations, deformations, and different viewpoints. In this paper, we develop a novel network named Gated Path Selection Network (GPSNet), which aims to adaptively select receptive fields while maintaining the dense sampling capability. In GPSNet, we first design a two-dimensional SuperNet, which densely incorporates features from growing receptive fields. And then, a Comparative Feature Aggregation (CFA) module is introduced to dynamically aggregate discriminative semantic context. In contrast to previous works that focus on optimizing sparse sampling locations on regular grids, GPSNet can adaptively harvest free form dense semantic context information. The derived adaptive receptive fields and dense sampling locations are data-dependent and flexible which can model various contexts of objects. On two representative semantic segmentation datasets, i.e., Cityscapes and ADE20K, we show that the proposed approach consistently outperforms previous methods without bells and whistles.

KW - Semantic segmentation

KW - adaptive context aggregation

KW - adaptive receptive fields and sampling locations

KW - local discriminative feature

UR - http://www.scopus.com/inward/record.url?scp=85099591781&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85099591781&partnerID=8YFLogxK

U2 - 10.1109/TIP.2020.3046921

DO - 10.1109/TIP.2020.3046921

M3 - Article

C2 - 33417546

AN - SCOPUS:85099591781

SN - 1057-7149

VL - 30

SP - 2436

EP - 2449

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

M1 - 9318517

ER -

Gated Path Selection Network for Semantic Segmentation

Abstract

Bibliographical note

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this