deformation @ tu

27
Computer Vision and Remote Sensing Technical University Berlin, Germany Department of Electrical Engineering IIT Roorkee, India Using Morphable Face Model to Improve Stereo Reconstruction and Visualising the Model on a Smartphone HARDIK JAIN Under the Guidance of Prof. Olaf HELLWICH Prof. RS ANAND

Upload: hardik-jain

Post on 28-Jan-2018

160 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Using Morphable Face Model to Improve Stereo Reconstruction and Visualising the

Model on a Smartphone

HARDIK JAINUnder the Guidance of

Prof. Olaf HELLWICH

Prof. RS ANAND

Page 2: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

PRESENTATION OUTLINE

Introduction

Motivation

Research Methodology

Results and Evaluation

Visualisation

Conclusion and Scope for Future Work

Further Reading

1 of 25

Page 3: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

INTRODUCTION

Face Reconstruction

Multiple Image ReconstructionVisual structure from motion

Stereo Reconstruction

Single Image Reconstruction3D Morphable Model

2 of 25

Page 4: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

3D Morphable Model

Statistical model of face meshes which are in dense correspondence.

Principal component analysis (PCA) is performed on these set of ๐‘€ meshes.

Mean shape าง๐‘ , ๐‘€ โˆ’ 1 Principal Components ๐‘†๐‘– and ๐œŽ๐‘ ,๐‘–2

eigen values

Shape parameter vector cs = ๐›ผ1, โ€ฆ , ๐›ผ๐‘€โˆ’1๐‘‡

iisi

M

i

model SsS 2

,

1

1=

=

INTRODUCTI

ONSingle Image Reconstruction

Morphable Model

3 of 25

Page 5: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

MOTIVATION Stereo Reconstruction

Stereo Model

Stereo Model

High Quality

Face Scan

Stereo Model and High Quality Scan Cloud

Compare

4 of 25

Page 6: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Face Model

MOTIVATION Single Image Reconstruction

Face Image Face Model

Face Model Cloud Compare

5 of 25

Page 7: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

MOTIVATION

Single Image

Reconstruction

Stereo Model

Deformed Face Model

Shape Information

Texture & Smoothness

Deformation

6 of 25

Page 8: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

RESEARCH METHODOLOGY

MorphableModel

Stereo Pair Image

Deformed Face Model

Single ImageLandmarking

Pose EstimationShape Fitting

Texture Extraction

Global & Local Deformation

Stereo Reconstruction

Face Model

7 of 25

Method Overview

Page 9: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Landmarking Annotation

To find ๐… = ๐‘ฅ1 ๐‘ฆ1, โ€ฆ , ๐‘ฅ๐‘› ๐‘ฆ๐‘›๐‘‡ โˆˆ โ„2๐‘›

Obtain Face Bounding Box

Cascade based Regression Method Initial estimate is centred to the Bounding box

Regressor rt ๐ผ, F(t)

F(t+1) = F(t) + rt ๐ผ, F(t)

RESEARCH

METHODOLOGYSingle Image Reconstruction

Landmark Annotated Face Image

8 of 25

Page 10: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Pose Estimation

Camera Orientation Matrix, ๐

๐ฑ๐’Š = ๐๐—๐’Š ,where ๐ฑ๐’Šโˆˆ โ„2 and ๐—๐’Šโˆˆ โ„3

Affine Camera Model

Gold Standard Algorithm of Hartley & Zisserman

RESEARCH

METHODOLOGYSingle Image Reconstruction

9 of 25

2D Landmark Points

3D Points on MM

Page 11: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Shape Fitting

Estimation of shape Parameter ๐œs

Probabilistic Approach

Minimise ๐ธ = ฯƒ๐‘–=13๐‘ ๐‘ฆ ๐‘š๐‘œ๐‘‘๐‘’๐‘™2๐ท,๐‘– โˆ’๐‘ฆ๐‘–

2

2 ๐œŽ 2๐ท,๐‘–2 + ๐œs

2

Face Image Face Shape Model

RESEARCH

METHODOLOGYSingle Image Reconstruction

iisi

M

i

model SsS 2

,

1

1=

=

10 of 25

Page 12: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Texture Extraction

Obtained from Face Image

Isomap Algorithm

Retaining Geodesic Distance

Texture Map

RESEARCH

METHODOLOGYSingle Image Reconstruction

Face Image

11 of 25

Page 13: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

STEREO RECONSTRUCTION

Obtain stereo pair images

Camera Calibration

Keypoint Detection

Triangulation

Dense Reconstruction

RESEARCH

METHODOLOGY

Stereo Model

12 of 25

Page 14: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

DEFORMATION

Motivation

Surface Registration

Approach

Two-step deformation

Global and Local transformation

๐’—๐‘— = ฮฆ๐‘™๐‘œ๐‘๐‘Ž๐‘™ โˆ˜ ฮฆ๐‘”๐‘™๐‘œ๐‘๐‘Ž๐‘™ (๐’—๐‘—)

RESEARCH

METHODOLOGY

13 of 25

Page 15: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Global Deformation

Using few control points

Radial Basis Function ๐œ™ ๐‘ฅ, ๐‘‹๐‘ = ๐œ™ ๐‘ฅ โˆ’ ๐‘‹๐‘

Weighted Combination of RBF

๐‘” ๐‘ฅ =

๐‘˜=1

๐‘

๐œ†๐‘ ๐œ™ ๐‘ฅ โˆ’ ๐‘‹๐‘

Gaussian Kernel ๐œ™๐‘–,๐‘ ๐‘ฅ = ๐‘’โˆ’

๐‘ฅ๐‘–โˆ’ ๐‘‹๐‘2

2๐œŽ2

RESEARCH

METHODOLOGYDeformation

14 of 25

Page 16: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

๐œ†๐‘ = ๐œ™โˆ’1๐‘” applied on Sourcepartial

๐œ†๐‘Ÿ,1โ‹ฎ

๐œ†๐‘Ÿ,๐‘

=

๐œ™1,1 โ‹ฏ ๐œ™1,๐‘โ‹ฎ โ‹ฑ โ‹ฎ

๐œ™๐‘‰,1 โ‹ฏ ๐œ™๐‘‰,๐‘

โˆ’1 ๐‘”๐‘Ÿ,1โ‹ฎ

๐‘”๐‘Ÿ,๐‘‰, for r = (x,y,z)

Face unspecific results for N=25

Few 100 milliseconds on intel quad core computer

RESEARCH

METHODOLOGYDeformation Global Deformation

Weight Coefficients (๐œ†๐‘)

15 of 25

Page 17: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Local Deformation

Smoothen the overshoots of RBF

Influence of nearby neighbors

Non-rigid transformation by k nearest neighbouring vertices

Procrustes Analysis

Affine transformation (๐๐‘– and ๐ญ๐‘–)

RESEARCH

METHODOLOGYDeformation

16 of 25

Page 18: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

ฮฆlocal ๐‘ฃ = w0 ๐‘ฃ ๐0๐‘ฃ + ๐ญ0 + ฯƒi=1๐พ wi ๐‘ฃ ๐i๐‘ฃ + ๐ญi

wi ๐‘ฃ =1

๐พ

dโˆ’ ๐‘ฃ๐‘–โˆ’ ๐‘ฃ

d, ๐‘‘ = ฯƒi=1

๐พ ๐‘ฃ๐‘– โˆ’ ๐‘ฃ

w0(๐‘ฃ) =1

๐พ

๐พ = 12

Requires Few seconds

RESEARCH

METHODOLOGYDEFORMATION

Local Deformation

1

5

4

2

3

6

17 of 25

Page 19: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

DEFORMED FACE MODEL

RESULTS AND

EVALUATION

Face Image Face Model Deformed Face Model

18 of 25

Page 20: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

EVALUATION

RESULTS AND

EVALUATION

19 of 25

3.5532 3.3073 2.8738

Page 21: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

EVALUATION

RESULTS AND

EVALUATION

20 of 25

4.2734 2.9985 2.0133

Page 22: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

EVALUATION

RESULTS AND

EVALUATION

21 of 25

2.9982 2.5212 2.174

Page 23: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

VISUALISATION

Virtual Reality

Using Smartphone and Cardboard Viewer

Android phones with support of OpenGL ES 3.1

Implemented on SDK provided by Google

22 of 25

Page 24: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

Smartphone View

VISUALISATION

23 of 25

Page 25: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

CONCLUSION AND SCOPE FOR FUTURE WORK

Improved reconstruction after information fusion

Technique Could be used for various other objects

Cheap alternative visualisation platform

Smartphone visualisation Environment Improvement

24 of 25

Page 26: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

FURTHER READING V. Blanz and T. Vetter, โ€œA morphable model for the synthesis of 3D faces,โ€ in Proceedings of the 26th annual conference on Computer graphics and interactive techniques. ACM Press/Addison-Wesley Publishing Co., 1999, pp. 187โ€“194.

P. Huber, G. Hu, R. Tena, P. Mortazavian, W. P. Koppen, W. Christmas, M. Ratsch, and J. Kittler, โ€œA multiresolution 3D morphable face model and fitting framework,โ€ in 11th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, February 2016.

P. Huber, Z.-H. Feng, W. Christmas, J. Kittler, and M. Rรคtsch, โ€œFitting 3D morphable models using local features,โ€ arXiv preprint arXiv:1503.02330, 2015.

V. Kazemi and J. Sullivan, โ€œOne millisecond face alignment with an ensemble of regression trees,โ€ in Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on. IEEE, 2014, pp. 1867โ€“ 1874.

R. W. Sumner, J. Schmid, and M. Pauly, โ€œEmbedded deformation for shape manipulation,โ€ ACM Transactions on Graphics (TOG), vol. 26, no. 3, p. 80, 2007.

โ€œCardboard.โ€ https://developers.google.com/cardboard/overview, 2016.

25 of 25

Page 27: Deformation @ TU

Computer Vision and Remote Sensing Technical University Berlin, Germany

Department of Electrical EngineeringIIT Roorkee, India

FURTHER READING

Thank You