A Machine-Learning Approach to Keypoint Detection and Landmarking on 3D Meshes

Clement Creusot; Nick Pears; Jim Austin

doi:10.1007/s11263-012-0605-9

A Machine-Learning Approach to Keypoint Detection and Landmarking on 3D Meshes

Clement Creusot, Nick Pears, Jim Austin

Computer Science

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

Abstract

We address the problem of automatically detecting a sparse set of 3D mesh vertices, likely to be good candidates for determining correspondences, even on soft organic objects. We focus on 3D face scans, on which single local shape descriptor responses are known to be weak, sparse or noisy. Our machine-learning approach consists of computing feature vectors containing $\dsn$ different local surface descriptors. These vectors are normalized with respect to the learnt distribution of those descriptors for some given target shape (landmark) of interest. Then, an optimal function of this vector is extracted that best separates this particular target shape from its surrounding region within the set of training data. We investigate two alternatives for this optimal function: a linear method, namely Linear Discriminant Analysis (LDA), and a non-linear method, namely AdaBoost. We evaluate our approach by landmarking 3D face scans in the FRGC v2 and Bosphorus 3D face datasets. Our system achieves state-of-the-art performance while being highly generic.

Original language	English
Pages (from-to)	146-179
Number of pages	34
Journal	International Journal of Computer Vision
Volume	102
Issue number	1-3
DOIs	https://doi.org/10.1007/s11263-012-0605-9
Publication status	Published - Mar 2012

Keywords

keypoint detection, landmarking, 3D face recognition, Machine Learning, LDA, AdaBoost

Access to Document

10.1007/s11263-012-0605-9

http://link.springer.com/content/pdf/10.1007%2Fs11263-012-0605-9.pdf

Cite this

@article{a65dbd59b3224c89be4483cd6b79ec98,

title = "A Machine-Learning Approach to Keypoint Detection and Landmarking on 3D Meshes",

abstract = "We address the problem of automatically detecting a sparse set of 3D mesh vertices, likely to be good candidates for determining correspondences, even on soft organic objects. We focus on 3D face scans, on which single local shape descriptor responses are known to be weak, sparse or noisy. Our machine-learning approach consists of computing feature vectors containing $\dsn$ different local surface descriptors. These vectors are normalized with respect to the learnt distribution of those descriptors for some given target shape (landmark) of interest. Then, an optimal function of this vector is extracted that best separates this particular target shape from its surrounding region within the set of training data. We investigate two alternatives for this optimal function: a linear method, namely Linear Discriminant Analysis (LDA), and a non-linear method, namely AdaBoost. We evaluate our approach by landmarking 3D face scans in the FRGC v2 and Bosphorus 3D face datasets. Our system achieves state-of-the-art performance while being highly generic. ",

keywords = "keypoint detection, landmarking, 3D face recognition, Machine Learning, LDA, AdaBoost",

author = "Clement Creusot and Nick Pears and Jim Austin",

year = "2012",

month = mar,

doi = "10.1007/s11263-012-0605-9",

language = "English",

volume = "102",

pages = "146--179",

journal = "International Journal of Computer Vision",

issn = "0920-5691",

publisher = "Springer",

number = "1-3",

}

TY - JOUR

T1 - A Machine-Learning Approach to Keypoint Detection and Landmarking on 3D Meshes

AU - Creusot, Clement

AU - Pears, Nick

AU - Austin, Jim

PY - 2012/3

Y1 - 2012/3

N2 - We address the problem of automatically detecting a sparse set of 3D mesh vertices, likely to be good candidates for determining correspondences, even on soft organic objects. We focus on 3D face scans, on which single local shape descriptor responses are known to be weak, sparse or noisy. Our machine-learning approach consists of computing feature vectors containing $\dsn$ different local surface descriptors. These vectors are normalized with respect to the learnt distribution of those descriptors for some given target shape (landmark) of interest. Then, an optimal function of this vector is extracted that best separates this particular target shape from its surrounding region within the set of training data. We investigate two alternatives for this optimal function: a linear method, namely Linear Discriminant Analysis (LDA), and a non-linear method, namely AdaBoost. We evaluate our approach by landmarking 3D face scans in the FRGC v2 and Bosphorus 3D face datasets. Our system achieves state-of-the-art performance while being highly generic.

AB - We address the problem of automatically detecting a sparse set of 3D mesh vertices, likely to be good candidates for determining correspondences, even on soft organic objects. We focus on 3D face scans, on which single local shape descriptor responses are known to be weak, sparse or noisy. Our machine-learning approach consists of computing feature vectors containing $\dsn$ different local surface descriptors. These vectors are normalized with respect to the learnt distribution of those descriptors for some given target shape (landmark) of interest. Then, an optimal function of this vector is extracted that best separates this particular target shape from its surrounding region within the set of training data. We investigate two alternatives for this optimal function: a linear method, namely Linear Discriminant Analysis (LDA), and a non-linear method, namely AdaBoost. We evaluate our approach by landmarking 3D face scans in the FRGC v2 and Bosphorus 3D face datasets. Our system achieves state-of-the-art performance while being highly generic.

KW - keypoint detection, landmarking, 3D face recognition, Machine Learning, LDA, AdaBoost

U2 - 10.1007/s11263-012-0605-9

DO - 10.1007/s11263-012-0605-9

M3 - Article

SN - 0920-5691

VL - 102

SP - 146

EP - 179

JO - International Journal of Computer Vision

JF - International Journal of Computer Vision

IS - 1-3

ER -