Presentation is loading. Please wait.

Presentation is loading. Please wait.

Mean-Field Theory and Its Applications In Computer Vision2

Similar presentations


Presentation on theme: "Mean-Field Theory and Its Applications In Computer Vision2"— Presentation transcript:

1 Mean-Field Theory and Its Applications In Computer Vision2

2 Dense CRF construction
Problem Formulation Grid CRF leads to over smoothing around boundaries Dense CRF is able to recover fine boundaries Grid CRF construction Dense CRF construction

3 Long Range Interaction
Able to recover proper flow for objects Teddy arms recovered using Global interaction Optical flow Optical flow and stereo reconstruction image Local interaction Global interaction Ground truth

4 Very Expensive Step (O(n2))
Marginal Update Marginal Update for large neighbourhood: Very Expensive Step (O(n2))

5 Inference in Dense CRF Time complexity increases Neighbourhood size
MCMC takes 36 hours on 50K variables Graph-cuts based algorithm takes hours

6 Inference in Dense CRF Time complexity increases Neighbourhood size
MCMC takes 36 hours on 50K variables Graph-cuts based algorithm takes hours Not practical for vision applications

7 Inference in Dense CRF Time complexity increases Neighbourhood size
MCMC takes 36 hours on 50K variables Graph-cuts based algorithm takes hours Filter-based Mean-field Inference takes 0.2 secs Possibility of development of many exciting vision applications

8 Efficient inference Assume Gaussian pairwise weight
Label compatibility function

9 Efficient inference Assume Gaussian pairwise weight
Mixture of Gaussians Spatial Bilateral

10 Bilateral filter output input output input reproduced from [Durand 02]

11 Marginal update Assume Gaussian pairwise weight

12 Very Expensive Step (O(n2))
How does it work Very Expensive Step (O(n2))

13 Message passing from all Xj to all Xi
Accumulates weights from all other pixels except itself

14 Message passing from all Xj to all Xi
Convert as Gaussian filtering step: Accumulate weights from all other pixels except itself

15 Message passing from all Xj to all Xi
Convert as Gaussian filtering step: Accumulate weights from all other pixels except itself

16 Efficient filtering steps
Now discuss how to do efficient filtering step


Download ppt "Mean-Field Theory and Its Applications In Computer Vision2"

Similar presentations


Ads by Google