Download presentation

Presentation is loading. Please wait.

Published byJaden Gonzales Modified over 4 years ago

1
Mean-Field Theory and Its Applications In Computer Vision4 1

2
Motivation 2 Helps in incorporating region/segment consistency in the model Pairwise CRF Higher order CRF

3
Motivation 3 Higher order terms can help in incorporating detectors into our model Image Without detector With detector

4
Marginal update 4 General form of meanfield update Expectation of the cost given variable v i takes a label

5
Marginal Update 5 General form of meanfield update Expectation of the clique given variable v i takes a label Summation over the possible states of the clique

6
Marginal Update in Meanfield 6 Some possible states Total number of possible states: 3 6 labels

7
Marginal Update in Meanfield 7 Exponential # of possible states for clique of size |c| and labels L: |L| C Expectation evaluation (summation) becomes infeasible

8
Marginal Update in Meanfield 8 Use restricted form of cost Pattern based potential

9
Marginal Update in Meanfield 9 Restrict the number of states to certain number of patterns Simple patterns Segment takes a label from label set of 4 patterns Or none

10
Marginal Update in Meanfield 10 Expectation calculation is quite efficient

11
Pattern based cost 11 Segment takes one of the forms

12
Pattern based cost 12 Segment does not take one of the forms

13
Pattern based cost 13 Simple patterns Pattern based higher order terms

14
P N Potts based patterns 14 P N Potts based patterns Potts patterns

15
Potts cost 15 Potts cost Potts patterns

16
Marginal Update in Meanfield 16 General form of meanfield update Expectation of the cost given variable v i takes a label

17
Expectation update 17 Probability of segment taking that label Potts patterns

18
Expectation update 18 Probability of segment not taking that label Potts patterns

19
Expectation update 19 Expectation update Potts patterns

20
Complexity 20 Expectation Updation: Time complexity O(NL) Preserves the complexity of original filter based method

21
PascalVOC-10 dataset 21 Inclusion of PN potts term: AlgorithmTime (s)OverallAv. RecallAv. I/U AHCRF+Cooc3681.4338.0130.09 Dense CRF0.6771.6334.5328.4 Dense + PN Potts 4.3579.8740.7130.18 Slight improvement in I/U score compared to more complex model which includes Pn Potts + cooccurrence terms Almost 8-9 times faster than the alpha-expansion based method

Similar presentations

OK

© Charles van Marrewijk, 2003 1 An Introduction to Geographical Economics Brakman, Garretsen, and Van Marrewijk.

© Charles van Marrewijk, 2003 1 An Introduction to Geographical Economics Brakman, Garretsen, and Van Marrewijk.

© 2018 SlidePlayer.com Inc.

All rights reserved.

To ensure the functioning of the site, we use **cookies**. We share information about your activities on the site with our partners and Google partners: social networks and companies engaged in advertising and web analytics. For more information, see the Privacy Policy and Google Privacy & Terms.
Your consent to our cookies if you continue to use this website.

Ads by Google

Ppt on my sweet home Ppt on indian politics quiz Ppt on construction company profile Ppt on schottky diode leakage Ppt on tata trucks specifications Ppt on pollution of air and water class 8 English 8 unit 13 read ppt on iphone Historical backgrounds for ppt on social media Ppt on structural changes in chromosomes ppt Ppt on forward rate agreement calculation