Facial landmarks localization using cascaded neural networks

Shahar Mahpod; Rig Das; Emanuele Maiorana; Yosi Keller; Patrizio Campisi

doi:10.1016/j.cviu.2021.103171

Facial landmarks localization using cascaded neural networks

Shahar Mahpod, Rig Das, Emanuele Maiorana, Yosi Keller, Patrizio Campisi

Roma Tre University

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

12 Scopus citations

Abstract

The accurate localization of facial landmarks is at the core of several face analysis tasks, such as face recognition and facial expression analysis, to name a few. In this work, we propose a novel localization approach based on a deep learning architecture that utilizes two paired cascaded subnetworks with convolutional neural network units. The cascaded units of the first subnetwork estimate heatmap-based encodings of the landmarks’ locations, while the cascaded units of the second subnetwork receive as inputs the outputs of the corresponding heatmap estimation units, and refine them through regression. The proposed scheme is experimentally shown to compare favorably with contemporary state-of-the-art schemes, especially when applied to images depicting challenging localization conditions.

Original language	English
Article number	103171
Journal	Computer Vision and Image Understanding
Volume	205
DOIs	https://doi.org/10.1016/j.cviu.2021.103171
State	Published - Apr 2021

Bibliographical note

Publisher Copyright:
© 2021 Elsevier Inc.

Funding

This work has been partially supported by COST Action 1206 {“De-identification for privacy protection in multimedia content”}. We gratefully acknowledge the support of NVIDIA Corporation for providing the Titan X Pascal GPU for this research work.

Funders	Funder number
NVIDIA
European Cooperation in Science and Technology

Keywords

Convolutional neural networks
Deep cascaded neural networks
Face alignment
Facial landmark localization

Access to Document

10.1016/j.cviu.2021.103171

Cite this

@article{e8f9ab04de39434a9e22003d45488c87,

title = "Facial landmarks localization using cascaded neural networks",

abstract = "The accurate localization of facial landmarks is at the core of several face analysis tasks, such as face recognition and facial expression analysis, to name a few. In this work, we propose a novel localization approach based on a deep learning architecture that utilizes two paired cascaded subnetworks with convolutional neural network units. The cascaded units of the first subnetwork estimate heatmap-based encodings of the landmarks{\textquoteright} locations, while the cascaded units of the second subnetwork receive as inputs the outputs of the corresponding heatmap estimation units, and refine them through regression. The proposed scheme is experimentally shown to compare favorably with contemporary state-of-the-art schemes, especially when applied to images depicting challenging localization conditions.",

keywords = "Convolutional neural networks, Deep cascaded neural networks, Face alignment, Facial landmark localization",

author = "Shahar Mahpod and Rig Das and Emanuele Maiorana and Yosi Keller and Patrizio Campisi",

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

year = "2021",

month = apr,

doi = "10.1016/j.cviu.2021.103171",

language = "אנגלית",

volume = "205",

journal = "Computer Vision and Image Understanding",

issn = "1077-3142",

publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Facial landmarks localization using cascaded neural networks

AU - Mahpod, Shahar

AU - Das, Rig

AU - Maiorana, Emanuele

AU - Keller, Yosi

AU - Campisi, Patrizio

PY - 2021/4

Y1 - 2021/4

N2 - The accurate localization of facial landmarks is at the core of several face analysis tasks, such as face recognition and facial expression analysis, to name a few. In this work, we propose a novel localization approach based on a deep learning architecture that utilizes two paired cascaded subnetworks with convolutional neural network units. The cascaded units of the first subnetwork estimate heatmap-based encodings of the landmarks’ locations, while the cascaded units of the second subnetwork receive as inputs the outputs of the corresponding heatmap estimation units, and refine them through regression. The proposed scheme is experimentally shown to compare favorably with contemporary state-of-the-art schemes, especially when applied to images depicting challenging localization conditions.

AB - The accurate localization of facial landmarks is at the core of several face analysis tasks, such as face recognition and facial expression analysis, to name a few. In this work, we propose a novel localization approach based on a deep learning architecture that utilizes two paired cascaded subnetworks with convolutional neural network units. The cascaded units of the first subnetwork estimate heatmap-based encodings of the landmarks’ locations, while the cascaded units of the second subnetwork receive as inputs the outputs of the corresponding heatmap estimation units, and refine them through regression. The proposed scheme is experimentally shown to compare favorably with contemporary state-of-the-art schemes, especially when applied to images depicting challenging localization conditions.

KW - Convolutional neural networks

KW - Deep cascaded neural networks

KW - Face alignment

KW - Facial landmark localization

UR - http://www.scopus.com/inward/record.url?scp=85101394505&partnerID=8YFLogxK

U2 - 10.1016/j.cviu.2021.103171

DO - 10.1016/j.cviu.2021.103171

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???

AN - SCOPUS:85101394505

SN - 1077-3142

VL - 205

JO - Computer Vision and Image Understanding

JF - Computer Vision and Image Understanding

M1 - 103171

ER -

Facial landmarks localization using cascaded neural networks

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this