Object recognition seminar S2006E01

2,925 views

Published on

Introduction

Published in: Technology, Economy & Finance
2 Comments
3 Likes
Statistics
Notes
No Downloads
Views
Total views
2,925
On SlideShare
0
From Embeds
0
Number of Embeds
93
Actions
Shares
0
Downloads
0
Comments
2
Likes
3
Embeds 0
No embeds

No notes for slide

Object recognition seminar S2006E01

  1. 1. Advanced seminar: Visual Object Recognition Dr. Lior Wolf School of Computer Science , Tel Aviv University
  2. 2. Credit: www . scholastic . com
  3. 3. Administrivia <ul><li>Mondays 13:00-15:00, unless.. </li></ul><ul><li>Dan David classrooms, room 106 </li></ul><ul><li>A 13 weeks seminar, unless.. </li></ul><ul><li>Grading: </li></ul><ul><ul><li>80% class presentation or class project </li></ul></ul><ul><ul><li>20% active participation in class </li></ul></ul><ul><li>Office hours: Wednesdays 13:00-15:00 </li></ul>
  4. 4. Example tasks: <ul><li>Street scene understanding </li></ul><ul><li>Face detection </li></ul><ul><li>Face identification </li></ul><ul><li>Image categorization </li></ul><ul><li>Pose estimation </li></ul><ul><li>Motion and behavior analysis </li></ul>
  5. 5. Street scene understanding Watch Out! Probably Hanging Out Credit: Stan Bileschi, CBCL
  6. 6. Face detection Credit: Intel Technology Journal, Volume 09, Issue 01
  7. 7. Face identification Credit: www . security - lab . com
  8. 8. Image categorization Images courtesy of Mobileye
  9. 9. Pose estimation Credit: Thomas Izo, CSAIL
  10. 10. Motion recognition Ballet turn: Input video: Output: The “Birmingham Royal Ballet” Credit: Eli Shechtman and Michal Irani
  11. 11. Basic image processing: <ul><li>Basic image operators </li></ul><ul><li>Template matching </li></ul><ul><li>Image transformations </li></ul>
  12. 12. An image is an array of pixels Credit: H. Joel Trussell, Eli Saber, and Michael Vrhel
  13. 13. Pixels can be manipulated I I>100 Conv2(I,[-1 1]) For more on image convolution, please google: “ image convolution gamedev”
  14. 14. Template matching Credit: F. Patin, gamedev.net
  15. 15. Fourier Transform Credit: R. Fisher, S. Perkins, A. Walker and E. Wolfart.
  16. 16. Major scheme: <ul><li>Image representation: features </li></ul><ul><li>Object learning: classification </li></ul>Credit: B. Heisele, Y. Ivanov, T. Poggio
  17. 17. Image features <ul><li>“ The most important element” (?) </li></ul><ul><li>Example: David Lowe’s SIFT descriptor </li></ul>Credit: David Lowe
  18. 18. Image features <ul><li>“ The most important element” (?) </li></ul><ul><li>Example: David Lowe’s SIFT descriptor </li></ul>Credit: David Lowe
  19. 19. Classification Credit: B. Heisele, Y. Ivanov, T. Poggio
  20. 20. Challenges: Too many to list..
  21. 21. Challenges 1: view point variation Michelangelo 1475-1564 slide by Fei Fei, Fergus & Torralba
  22. 22. Challenges 2: illumination slide credit: S. Ullman
  23. 23. Challenges 3: occlusion Magritte, 1957 slide by Fei Fei, Fergus & Torralba
  24. 24. Challenges 4: scale slide by Fei Fei, Fergus & Torralba
  25. 25. Challenges 5: deformation Xu, Beihong 1943 slide by Fei Fei, Fergus & Torralba
  26. 26. Challenges 6: background clutter Klimt, 1913 slide by Fei Fei, Fergus & Torralba
  27. 27. Challenges 7: object intra-class variation slide by Fei-Fei, Fergus & Torralba
  28. 28. Challenges 8: local ambiguity slide by Fei-Fei, Fergus & Torralba
  29. 29. Challenges 9: the world behind the image Credit: A. Efros

×