1. Automatic User Interaction Correction via Multi-label Graph-cuts Antonio Hernández-Vela, Carlos Primo and Sergio Escalera Workshop on Human Interaction in Computer Vision 12 nd November 2011

2. Semi-automatic segmentation 2

3. 3

4. Outline 1.Introduction 2.Methodology 3.Experiments & results 4.Conclusion 4

5. Introduction How much interaction is required from the user? What if the user makes a mistake? Automatic correction Multi-label approach 5 X X

6. Outline 1.Introduction 2.Methodology 3.Experiments & results 4.Conclusion 6

7. Multi-label graph-cuts 1 7 Energy minimization framework Image restorationStereo reconstruction Image segmentation 1 Y. Boykov, O. Veksler, and R. Zabih. Fast approximate energy minimization via graph cuts. PAMI, nov 2001.

8. Segmentation with Multi-label graph-cuts 8 1.User initialization 2.Unary potential 3.Pairwise potential 4.Segmentation (α-expansion)

9. Likelihood-based correction 9 Color model: Gaussian Mixture Model :Seeds for label : weight of GMM component

10. Likelihood-based correction: example 10

11. Likelihood-based correction: example 11

12. Outline 1.Introduction 2.Methodology 3.Experiments & results 4.Conclusion 12

13. Experiments: Data 13 Human Limb DB 2 : 227 images from 25 people, GT provided Random selection of 10 images –from 10 different people– 2 http://www.maia.ub.es/~sergio/linked/humanlimbdb.zip

14. Experiments: Setup 14 Random interaction simulation Select 0.2% of total pixels from each label Repeat the process 10 times 4 scenarios: With/Without error Single/Combined mode

15. Results without errors 15

16. Results with 25% of error 16

17. Results with 45% of error 17

18. Qualitatve Results (single mode) 18

19. Qualitatve Results (combined mode) 19

20. Outline 1.Introduction 2.Methodology 3.Experiments & results 4.Conclusion 20

21. Conclusion 21 Methodology for automatic user interaction correction in multi-label graph-cuts image segmentation. Study on the influence of the amount of interaction from the user. –With and without errors. Future work Further experiment using real user interactions. Adapt the algorithm to the user.

22. Automatic User Interaction Correction via Multi-label Graph-cuts Antonio Hernández-Vela, Carlos Primo and Sergio Escalera Workshop on Human Interaction in Computer Vision 12 nd November 2011 Thank You! Questions?