A Human Ear Reconstruction Autoencoder: HERA

Hao Sun; Nicholas Edwin Pears; Hang Dai

A Human Ear Reconstruction Autoencoder: HERA

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

The ear, as an important part of the human head, has received much less attention compared to the human face in the area of computer vision. Inspired by previous work on monocular 3D face reconstruction using an autoencoder structure to achieve self-supervised learning, we aim to utilise such a framework to tackle the 3D ear reconstruction task, where more subtle and difficult curves and features are present on the 2D ear input images. Our Human Ear Reconstruction Autoencoder (HERA) system predicts 3D ear poses and shape parameters for 3D ear meshes, without any supervision to these parameters. To make our approach cover the variance for in-the-wild images, even grayscale images, we propose an in-the-wild ear colour model. The constructed end-to-end self-supervised model is then evaluated both with 2D landmark localisation performance and the appearance of the reconstructed 3D ears.

Original language	English
Title of host publication	16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications
Subtitle of host publication	VISIGRAPP 2021
Publisher	SciTePress
Pages	136-145
Number of pages	10
Volume	5
ISBN (Electronic)	9789897584886
Publication status	Published - 8 Feb 2021
Event	16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications. - Online Duration: 8 Feb 2021 → 10 Feb 2021 http://www.visapp.visigrapp.org/Home.aspx

Conference

Conference	16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications.
Abbreviated title	VISAPP 2021
Period	8/02/21 → 10/02/21
Internet address	http://www.visapp.visigrapp.org/Home.aspx

Keywords

Ear, 3D ear model, 3D morphable model, 3D reconstruction, Self-supervised learning, Autoencoder

Cite this

@inproceedings{68da790dc53448aa95d4235bfee44ca0,

title = "A Human Ear Reconstruction Autoencoder: HERA",

abstract = "The ear, as an important part of the human head, has received much less attention compared to the human face in the area of computer vision. Inspired by previous work on monocular 3D face reconstruction using an autoencoder structure to achieve self-supervised learning, we aim to utilise such a framework to tackle the 3D ear reconstruction task, where more subtle and difficult curves and features are present on the 2D ear input images. Our Human Ear Reconstruction Autoencoder (HERA) system predicts 3D ear poses and shape parameters for 3D ear meshes, without any supervision to these parameters. To make our approach cover the variance for in-the-wild images, even grayscale images, we propose an in-the-wild ear colour model. The constructed end-to-end self-supervised model is then evaluated both with 2D landmark localisation performance and the appearance of the reconstructed 3D ears.",

keywords = "Ear, 3D ear model, 3D morphable model, 3D reconstruction, Self-supervised learning, Autoencoder",

author = "Hao Sun and Pears, {Nicholas Edwin} and Hang Dai",

year = "2021",

month = feb,

day = "8",

language = "English",

volume = "5",

pages = "136--145",

booktitle = "16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications",

publisher = "SciTePress",

note = "16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications., VISAPP 2021 ; Conference date: 08-02-2021 Through 10-02-2021",

url = "http://www.visapp.visigrapp.org/Home.aspx",

}

TY - GEN

T1 - A Human Ear Reconstruction Autoencoder

T2 - 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications.

AU - Sun, Hao

AU - Pears, Nicholas Edwin

AU - Dai, Hang

PY - 2021/2/8

Y1 - 2021/2/8

N2 - The ear, as an important part of the human head, has received much less attention compared to the human face in the area of computer vision. Inspired by previous work on monocular 3D face reconstruction using an autoencoder structure to achieve self-supervised learning, we aim to utilise such a framework to tackle the 3D ear reconstruction task, where more subtle and difficult curves and features are present on the 2D ear input images. Our Human Ear Reconstruction Autoencoder (HERA) system predicts 3D ear poses and shape parameters for 3D ear meshes, without any supervision to these parameters. To make our approach cover the variance for in-the-wild images, even grayscale images, we propose an in-the-wild ear colour model. The constructed end-to-end self-supervised model is then evaluated both with 2D landmark localisation performance and the appearance of the reconstructed 3D ears.

AB - The ear, as an important part of the human head, has received much less attention compared to the human face in the area of computer vision. Inspired by previous work on monocular 3D face reconstruction using an autoencoder structure to achieve self-supervised learning, we aim to utilise such a framework to tackle the 3D ear reconstruction task, where more subtle and difficult curves and features are present on the 2D ear input images. Our Human Ear Reconstruction Autoencoder (HERA) system predicts 3D ear poses and shape parameters for 3D ear meshes, without any supervision to these parameters. To make our approach cover the variance for in-the-wild images, even grayscale images, we propose an in-the-wild ear colour model. The constructed end-to-end self-supervised model is then evaluated both with 2D landmark localisation performance and the appearance of the reconstructed 3D ears.

KW - Ear, 3D ear model, 3D morphable model, 3D reconstruction, Self-supervised learning, Autoencoder

M3 - Conference contribution

VL - 5

SP - 136

EP - 145

BT - 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications

PB - SciTePress

Y2 - 8 February 2021 through 10 February 2021

ER -