cs381v paper presentation
play

CS381V Paper Presentation Chun-Chen Kuo Selective Search for - PowerPoint PPT Presentation

CS381V Paper Presentation Chun-Chen Kuo Selective Search for Object Recognition Outline Problem statement Technical details Evaluation Extensions Problem Statement Image Classification Input: training set and


  1. CS381V Paper Presentation Chun-Chen Kuo

  2. Selective Search for Object Recognition

  3. Outline • Problem statement • Technical details • Evaluation • Extensions

  4. Problem Statement

  5. Image Classification • Input: training set and test set • Output: the class of the test images and the confidence scores ML Car (0.8) model image from ImageNet

  6. Object Detection • Input: training set 
 and a test set • Output: all objects in the test images and 
 their bounding boxes ML Car (0.9) model image from ImageNet

  7. • How to turn object detection problem to image classification problem? image from ImageNet ML Car (0.9) model

  8. • How many sub-regions should we test and how do we generate them? image from ImageNet ML Car (0.9) model

  9. Exhaustive Search • Generate all possible windows • Complexity: all size all location image from ImageNet

  10. Selective Search • Reduce the number of hypotheses while keep recall high • Select some high quality hypotheses, which are subset of all possible hypotheses

  11. Technical Details

  12. Intuition • Explore image structure and group regions from small scale to high scale ( hierarchical grouping) Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  13. Algorithm Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  14. Similarity Function Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  15. Similarity Function Color Texture Part Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  16. Similarity Function • Color? Texture? Part? • No single strategy to group regions • Need to diversify by using complementary similarity measures

  17. Similarity Function • Color similarity: • Normalized color histogram with 25 bins: • Propagate through the hierarchy: Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  18. Similarity Function • Texture similarity: • Take Gaussian derivatives in 8 orientations, and extract histogram with bin size=10: Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  19. Similarity Function • Size similarity: • Merge small regions first 
 Prevent a big region eating small regions

  20. Similarity Function • Fill similarity: Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  21. Similarity Function • Combine them: a=[1,1,1,1] => C+T+S+F a=[0,1,1,1] => T+S+F Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  22. Complementary Color Space • Also diversify in color space invariance Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  23. Evaluation

  24. 
 
 
 
 Metrics • Average Best Overlap (ABO) 
 • Mean Average Best Overlap (MABO) 
 mean of ABO over all classes

  25. Some Examples Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  26. Flat v.s Hierarchy Hierarchy is good! Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  27. Diversification Strategies Diversification is good! Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  28. Compare to Other Methods State of the art! Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  29. Contribution and Strength • Hierarchical grouping and diversification strategies • Nice trade-off between quality(MABO) and quantity(# window)

  30. Weakness • The algorithm for sorting the object hypotheses s.t. the most likely hypothesis comes first • No evaluation on it? • Favor large scale but times rand() to prevent over- favor?

  31. Extension

  32. R-CNN • Regions with Convolutional Neural Network features Rich feature hierarchies for accurate object detection and semantic segmentation. R. Girshick et al. CVPR 2013

  33. Visual-Semantic Alignment A. Karpathy and L. Fei-Fei. Deep visual-semantic alignments for generating image descriptions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3128–3137, 2015.

  34. Reference • J. R. Uijlings, K. E. van de Sande, T. Gevers, and A. W. Smeulders. Selective search for object recognition. International journal of computer vision, 104(2):154–171, 2013 • R. Girshick, J. Donahue, T. Darrell, and J. Malik. Rich feature hierarchies for accurate object detection and semantic segmentation. In CVPR, 2014. • A. Karpathy and L. Fei-Fei. Deep visual-semantic alignments for generating image descriptions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3128–3137, 2015. 


  35. Appendix

  36. Application on Object Detection Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  37. Diversification Strategies Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

  38. Trade-off between Quality and Quantity Selective Search for Object Recognition. J. Uijilings, K. van de Sande, T. Gevers, A. Smeulders. IJCV 2013

Recommend


More recommend