Evidential modeling for pose estimation

1
Evidential modeling for pose estimation

• Fabio Cuzzolin, Ruggero Frezza
• Computer Science Department
• UCLA

2
Myself

• Masters thesis on gesture recognition
• at the University of Padova
• Ph.D. thesis on the theory of evidence
• Post-doc in Milan with the Image and Sound
  Processing group
• Post-doc at UCLA in the Vision Lab

3
My past work

• geometric approach to the theory of belief
  functions
• space of belief functions
• geometry of Dempsters rule

4
.. again ..

• algebra of compatible frames
• linear independence on lattices
• action recognition and object tracking
• metrics on the space of dynamical models

5
and todays talk
1

• the pose estimation problem

2

• model-free pose estimation

3

• evidential model

4

• experimental results

6
Pose estimation

• estimating the pose (internal configuration) of
  a moving body from the available images

CAMERA
t0
tT
7
Model-based estimation

• if you have an a-priori model of the object ..
• .. you can exploit it to help (or drive) the
  estimation

8
Model-free estimation

• if you do not have any information about the
  body..

9
Collecting training data

• motion capture system

10
Training data

• when the object performs some significant
  movements in front of the camera
• a finite collection of configuration values are
  provided by the motion capture system

q
q
1
T
y
y
1
T

• while a sequence of features is computed from
  the image(s)

11
Learning feature-pose maps

• Hidden Markov models provide a way to build
• feature-pose maps from the training data

12
Evidential model

• approximate feature spaces ..

• .. and approximate parameter space ..

13
Estimation
14
Human body tracking

• two experiments, two views

15
Feature extraction

• three steps original image, color segmentation,
  bounding box

16
Performances

• comparison of three models left view only, right
  view only, both views

17
Estimation errors

• Euclidean distance between real and predicted
  marker position

18
Visual estimate
19
Conclusions

• pose estimation of unknown objects is a difficult
  task
• a bottom-up model has to be built from the data
  in a training session
• the DS framework allows to formalize the idea of
  feature-pose maps in a natural way through the
  notion of compatible frames
• Dempsters combination provides a method to
  integrate features to increase robustness