close

Вход

Забыли?

вход по аккаунту

?

Strong Priors for Multi view Reconstruction

код для вставкиСкачать
Using Strong Shape Priors for
Multiview Reconstruction
Yunda Sun
Pushmeet Kohli
Mathieu Bray Philip HS Torr
Department of Computing
Oxford Brookes University
Objective
Images
Silhouettes
+
Parametric
Model
Pose
Estimate
Reconstruction
[Images Courtesy: M. Black, L. Sigal]
Outline
Multi-view Reconstruction
пЃ® Shape Models as Strong Priors
пЃ® Object Specific MRF
пЃ® Pose Estimation
пЃ® Results
пЃ®
Outline
Multi-view Reconstruction
пЃ® Shape Models as Strong Priors
пЃ® Object Specific MRF
пЃ® Pose Estimation
пЃ® Results
пЃ®
Multiview Reconstruction
Need for Shape
Priors
Multiview Reconstruction
пЃ®
No Priors
• Silhouette Intersection
• Space Carving
пЃ®
Weak Priors
• Surface smoothness
– Snow et al. CVPR ’00
• Photo consistency and
smoothness
– Kolmogorov and Zabih
[ECCV ’02]
– Vogiatzis et al. [CVPR ’05]
[Image Courtesy: Vogiatzis et al.]
Outline
Multi-view Reconstruction
пЃ® Shape Models as Strong Priors
пЃ® Object Specific MRF
пЃ® Pose Estimation
пЃ® Results
пЃ®
Shape-Priors for Segmentation
пЃ®
OBJ-CUT [Kumar et al., CVPR ’05]
• Integrate Shape Priors in a MRF
пЃ®
POSE-CUT [Bray et al., ECCV ’06]
• Efficient Inference of Model Parameters
Parametric Object Models as Strong
Priors
пЃ®
Layered Pictorial Structures
пЃ®
Articulated Models
пЃ®
Deformable Models
Outline
Multi-view Reconstruction
пЃ® Shape Models as Strong Priors
пЃ® Object Specific MRF
пЃ® Pose Estimation and Reconstruction
пЃ® Results
пЃ®
Object-Specific MRF
Object-Specific MRF
Energy Function
Shape
Prior
Unary
Likelihood
Smoothness
Prior
x : Voxel label
Оё : Model Shape
Object-Specific MRF
Shape Prior
: shortest distance of voxel i from the rendered model
x : Voxel label
Оё : Model Shape
Object-Specific MRF
Smoothness Prior
Potts Model
x : Voxel label
Оё : Model Shape
Object-Specific MRF
Unary Likelihood
For a soft constraint we use a large
constant K instead of infinity
x : Voxel label
Оё : Model Shape
: Visual Hull
Object-Specific MRF
Energy Function
Shape
Prior
Unary
Likelihood
Smoothness
Prior
Can be solved using
Graph cuts
[Kolmogorov and Zabih,
ECCV02 ]
Object-Specific MRF
Energy Function
Shape
Prior
Unary
Likelihood
Smoothness
Prior
How to find the optimal
Pose?
Outline
Multi-view Reconstruction
пЃ® Shape Models as Strong Priors
пЃ® Object Specific MRF
пЃ® Pose Estimation
пЃ® Results
пЃ®
Inference of Pose Parameters
Rotation and Translation of
Torso in X axes
Rotation of left shoulder in X
and Z axes
Inference of Pose Parameters
Let F(У©) =
Minimize F(У©) using Powell Minimization
Computational Problem:
Each evaluation of F(У©) requires a graph cut to be
computed. (computationally expensive!!) BUT..
Solution: Use the dynamic graph cut algorithm
[Kohli&Torr, ICCV 2005]
Outline
Multi-view Reconstruction
пЃ® Shape Models as Strong Priors
пЃ® Object Specific MRF
пЃ® Pose Estimation
пЃ® Results
пЃ®
Experiments
пЃ®
Deformable Models
пЃ®
Articulated Models
• Reconstruction Results
• Human Pose Estimation
Deformable Models
Four Cameras
пЃ® 1.5 x 105 voxels
пЃ® DOF of Model: 5
пЃ®
Visual Hull
Our Reconstruction
Shape Model
Articulated Models
Articulated Models
Four Cameras
пЃ® 106 voxels
пЃ® DOF of Model: 26
пЃ®
Shape Model
Camera Setup
Articulated Models
Let F(У©) =
500 function evaluations of F(Оё) required
пЃ® Time per evaluation: 0.15 sec
пЃ® Total time: 75 sec
пЃ®
Articulated Models
Visual
Hull
Our
Reconstruction
Pose Estimation Results
Visual Hull
Reconstruction
Pose Estimate
Pose Estimation Results
пЃ®
Quantitative Results
• 6 uniformly distributed cameras
• 12 degree (RMS) error over 21 joint angles
Pose Estimation Results
пЃ®
Qualitative Results
Pose Estimation Results
Video 1, Camera 1
Pose Estimation Results
Video 1, Camera 2
Pose Estimation Results
Video 2, Camera 1
Pose Estimation Results
Video 2, Camera 2
Future Work
• Use dimensionality reduction to reduce the number
of pose parameters.
- results in less number of pose parameters to
optimize
- would speed up inference
• High resolution reconstruction by a coarse to fine
strategy
• Parameter Learning in Object Specific MRF
Thank You
Object-Specific MRF
Energy Function
Shape
Prior
Unary
Likelihood
+
Smoothness
Prior
Документ
Категория
Презентации
Просмотров
7
Размер файла
7 636 Кб
Теги
1/--страниц
Пожаловаться на содержимое документа