Bottom-Up Attention Guidance for Recurrent Image Recognition

Hamed Rezazadegan Tavakoli; Ali Borji; Rao Anwer; Esa Rahtu; Juho Kannala

doi:10.1109/ICIP.2018.8451537

Bottom-Up Attention Guidance for Recurrent Image Recognition

Hamed Rezazadegan Tavakoli
, Ali Borji
, Rao Anwer
, Esa Rahtu
, Juho Kannala

Tampere University of Technology
University of Central Florida

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

4 Citations (Scopus)

Abstract

This paper presents a recurrent neural network architecture, guided by the bottom-up attention, for the recognition task. The proposed architecture processes an input image as a sequence of selectively chosen patches. The patches are chosen from the salient regions of the input image. Using human driven saliency maps from gaze, the benefit of such a selection process is first shown. Next, the performance of computational models of bottom-up attention are assessed as alternative to human attention.

Original language	English
Title of host publication	2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings
Publisher	IEEE
Pages	3004-3008
Number of pages	5
ISBN (Electronic)	9781479970612
DOIs	https://doi.org/10.1109/ICIP.2018.8451537
Publication status	Published - 2018
MoE publication type	A4 Conference publication
Event	IEEE International Conference on Image Processing - Athens, Greece Duration: 7 Oct 2018 → 10 Oct 2018 Conference number: 25

Conference

Conference	IEEE International Conference on Image Processing
Abbreviated title	ICIP
Country/Territory	Greece
City	Athens
Period	07/10/2018 → 10/10/2018

Keywords

Recurrent neural networks
deep neural networks
gaze
saliency
image recognition

Access to Document

10.1109/ICIP.2018.8451537

Science-IT
Hakala, M. (Manager)
School of Science
Facility/equipment: Facility

Cite this

@inproceedings{9a883dc562aa46bf9a17e2d53383a341,

title = "Bottom-Up Attention Guidance for Recurrent Image Recognition",

abstract = "This paper presents a recurrent neural network architecture, guided by the bottom-up attention, for the recognition task. The proposed architecture processes an input image as a sequence of selectively chosen patches. The patches are chosen from the salient regions of the input image. Using human driven saliency maps from gaze, the benefit of such a selection process is first shown. Next, the performance of computational models of bottom-up attention are assessed as alternative to human attention.",

keywords = "Recurrent neural networks, deep neural networks, gaze, saliency, image recognition",

author = "\{Rezazadegan Tavakoli\}, Hamed and Ali Borji and Rao Anwer and Esa Rahtu and Juho Kannala",

year = "2018",

doi = "10.1109/ICIP.2018.8451537",

language = "English",

pages = "3004--3008",

booktitle = "2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings",

publisher = "IEEE",

address = "United States",

note = "IEEE International Conference on Image Processing, ICIP ; Conference date: 07-10-2018 Through 10-10-2018",

}

Rezazadegan Tavakoli, H, Borji, A, Anwer, R, Rahtu, E & Kannala, J 2018, Bottom-Up Attention Guidance for Recurrent Image Recognition. in 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings., 8451537, IEEE, pp. 3004-3008, IEEE International Conference on Image Processing, Athens, Greece, 07/10/2018. https://doi.org/10.1109/ICIP.2018.8451537

TY - GEN

T1 - Bottom-Up Attention Guidance for Recurrent Image Recognition

AU - Rezazadegan Tavakoli, Hamed

AU - Borji, Ali

AU - Anwer, Rao

AU - Rahtu, Esa

AU - Kannala, Juho

N1 - Conference code: 25

PY - 2018

Y1 - 2018

N2 - This paper presents a recurrent neural network architecture, guided by the bottom-up attention, for the recognition task. The proposed architecture processes an input image as a sequence of selectively chosen patches. The patches are chosen from the salient regions of the input image. Using human driven saliency maps from gaze, the benefit of such a selection process is first shown. Next, the performance of computational models of bottom-up attention are assessed as alternative to human attention.

AB - This paper presents a recurrent neural network architecture, guided by the bottom-up attention, for the recognition task. The proposed architecture processes an input image as a sequence of selectively chosen patches. The patches are chosen from the salient regions of the input image. Using human driven saliency maps from gaze, the benefit of such a selection process is first shown. Next, the performance of computational models of bottom-up attention are assessed as alternative to human attention.

KW - Recurrent neural networks

KW - deep neural networks

KW - gaze

KW - saliency

KW - image recognition

U2 - 10.1109/ICIP.2018.8451537

DO - 10.1109/ICIP.2018.8451537

M3 - Conference article in proceedings

SP - 3004

EP - 3008

BT - 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings

PB - IEEE

T2 - IEEE International Conference on Image Processing

Y2 - 7 October 2018 through 10 October 2018

ER -

Bottom-Up Attention Guidance for Recurrent Image Recognition

Abstract

Conference

Keywords

Access to Document

Fingerprint

Equipment

Science-IT

Cite this