Presentation is loading. Please wait.

Presentation is loading. Please wait.

3D Human Body Pose Estimation using GP-LVM Moin Nabi Computer Vision Group Institute for Research in Fundamental Sciences (IPM)

Similar presentations


Presentation on theme: "3D Human Body Pose Estimation using GP-LVM Moin Nabi Computer Vision Group Institute for Research in Fundamental Sciences (IPM)"— Presentation transcript:

1 3D Human Body Pose Estimation using GP-LVM Moin Nabi Computer Vision Group Institute for Research in Fundamental Sciences (IPM)

2 Introduction to Human Pose Estimation Articulated pose estimation from single-view monocular image(s)

3 Application of Human Pose Estimation ■ Entertainment: Animation, Games ■ Security: Surveillance ■ Understanding: Gesture/Activity recognition

4 Difficulties of Human Pose estimation ■ Appearance/size/shape of people can vary dramatically ■ The bones and joints are observable indirectly (obstructed by clothing) ■ Occlusions ■ High dimensionality of the state space ■ Lose of depth information in 2D image projections

5 Difficulties of Human Pose estimation ■ Challenging Human Motion

6 Problem Backgrounds ■ Pose Estimation From Monocular Image Goal: Reliable 3D Human Pose Estimation from single-camera input

7 Gaussian process

8

9

10

11 a 5x5 covariance matrix and a 3-d input vector was used to calculate the 2-d output mean vector and the corresponding variances

12 Gaussian process Use for Regression

13 Linear Dimension Reduction

14 Find the best latent inputs by maximizing the marginal likelihood under the constraint that all visible variables must share the same latent values.

15 Linear Dimension Reduction Find the best latent inputs by maximizing the marginal likelihood under the constraint that all visible variables must share the same latent values.

16 Linear Dimension Reduction

17

18

19 Gaussian process

20 Nonlinear Dimension Reduction

21

22

23

24

25

26 Human Pose Estimation using GP-LVM Image -> Pose In Latent Space

27 Human Pose Estimation using GP-LVM Motion capture example, representing 102-D data in 2-D

28 Human Pose Estimation using GP-LVM

29 Result

30

31 Pose from Action

32 Thank You Pose from Action

33 Different Action has Different shape in latent space Future Work Guess Action from shape of model in latent space

34 Thank You


Download ppt "3D Human Body Pose Estimation using GP-LVM Moin Nabi Computer Vision Group Institute for Research in Fundamental Sciences (IPM)"

Similar presentations


Ads by Google