Reflective Field for Pixel-Level Tasks

Liang Zhang; X. Kong; Peiyi Shen; Guangming Zhu; Syed Shah; Mohammed Bennamoun; J. Song

doi:10.1109/ICPR.2018.8545817

Reflective Field for Pixel-Level Tasks

Liang Zhang, X. Kong, Peiyi Shen, Guangming Zhu, Syed Shah, Mohammed Bennamoun, J. Song

School of Physics, Mathematics and Computing

Research output: Chapter in Book/Conference paper › Conference paper

Abstract

PixelNet has achieved great success in dense prediction
problems with a pure pixel-level architecture, but there is
still much room for improvement. In this paper, we start from
PixelNet and discuss the pixel-level architecture called hypercolumn
and its limitations in building feature representation with
rich semantic information. To achieve this goal, we propose a
concept in the context of neural networks called reflective field,
representing the area reflected by the origin input. Furthermore,
the proposed reflective field is used to solve the limitations of
the hypercolumn architecture. Specifically, we give the method of
calculating the size of the reflective field and analyze the effective
reflective field in the calculated area. Then, we use the reflective
field to build a new hypercolumn architecture, which has a more
rational construction. The results on PASCAL VOC segmentation
dataset with our new architecture are improved.

Original language	English
Title of host publication	Reflective Field for Pixel-Level Tasks
Publisher	IEEE, Institute of Electrical and Electronics Engineers
Pages	529-534
ISBN (Print)	9781538637883
DOIs	https://doi.org/10.1109/ICPR.2018.8545817
Publication status	Published - 10 Oct 2018

Access to Document

10.1109/ICPR.2018.8545817

Fingerprint Dive into the research topics of 'Reflective Field for Pixel-Level Tasks'. Together they form a unique fingerprint.

Cite this

@inproceedings{69dadadcb1a349168acff1db27eb918f,

title = "Reflective Field for Pixel-Level Tasks",

abstract = "PixelNet has achieved great success in dense predictionproblems with a pure pixel-level architecture, but there isstill much room for improvement. In this paper, we start fromPixelNet and discuss the pixel-level architecture called hypercolumnand its limitations in building feature representation withrich semantic information. To achieve this goal, we propose aconcept in the context of neural networks called reflective field,representing the area reflected by the origin input. Furthermore,the proposed reflective field is used to solve the limitations ofthe hypercolumn architecture. Specifically, we give the method ofcalculating the size of the reflective field and analyze the effectivereflective field in the calculated area. Then, we use the reflectivefield to build a new hypercolumn architecture, which has a morerational construction. The results on PASCAL VOC segmentationdataset with our new architecture are improved.",

author = "Liang Zhang and X. Kong and Peiyi Shen and Guangming Zhu and Syed Shah and Mohammed Bennamoun and J. Song",

year = "2018",

month = oct,

day = "10",

doi = "10.1109/ICPR.2018.8545817",

language = "English",

isbn = "9781538637883",

pages = "529--534",

booktitle = "Reflective Field for Pixel-Level Tasks",

publisher = "IEEE, Institute of Electrical and Electronics Engineers",

address = "United States",

}

TY - GEN

T1 - Reflective Field for Pixel-Level Tasks

AU - Zhang, Liang

AU - Kong, X.

AU - Shen, Peiyi

AU - Zhu, Guangming

AU - Shah, Syed

AU - Bennamoun, Mohammed

AU - Song, J.

PY - 2018/10/10

Y1 - 2018/10/10

N2 - PixelNet has achieved great success in dense predictionproblems with a pure pixel-level architecture, but there isstill much room for improvement. In this paper, we start fromPixelNet and discuss the pixel-level architecture called hypercolumnand its limitations in building feature representation withrich semantic information. To achieve this goal, we propose aconcept in the context of neural networks called reflective field,representing the area reflected by the origin input. Furthermore,the proposed reflective field is used to solve the limitations ofthe hypercolumn architecture. Specifically, we give the method ofcalculating the size of the reflective field and analyze the effectivereflective field in the calculated area. Then, we use the reflectivefield to build a new hypercolumn architecture, which has a morerational construction. The results on PASCAL VOC segmentationdataset with our new architecture are improved.

AB - PixelNet has achieved great success in dense predictionproblems with a pure pixel-level architecture, but there isstill much room for improvement. In this paper, we start fromPixelNet and discuss the pixel-level architecture called hypercolumnand its limitations in building feature representation withrich semantic information. To achieve this goal, we propose aconcept in the context of neural networks called reflective field,representing the area reflected by the origin input. Furthermore,the proposed reflective field is used to solve the limitations ofthe hypercolumn architecture. Specifically, we give the method ofcalculating the size of the reflective field and analyze the effectivereflective field in the calculated area. Then, we use the reflectivefield to build a new hypercolumn architecture, which has a morerational construction. The results on PASCAL VOC segmentationdataset with our new architecture are improved.

U2 - 10.1109/ICPR.2018.8545817

DO - 10.1109/ICPR.2018.8545817

M3 - Conference paper

SN - 9781538637883

SP - 529

EP - 534

BT - Reflective Field for Pixel-Level Tasks

PB - IEEE, Institute of Electrical and Electronics Engineers

ER -