Understanding the Importance of Different Edge Types in Image Recognition

This content delves into the significance of various edge types in image recognition, exploring the impact of edges such as albedo, depth, surface normal, and shadow on the recognition process. Researchers have studied the effects of different edge types through works dating back to the 1960s, aiming to isolate their specific contributions. Synthetic scenes have been utilized to analyze edges' roles in different visual contexts, shedding light on their implications for image processing and scene understanding.

• Image recognition
• Edge types
• Albedo
• Depth
• Synthetic scenes

marianna Follow

Uploaded on Jul 19, 2024 | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

Presentation Transcript

1. Which edges matter? Aayush Bansal (CMU), Adarsh Kowdle (Microsoft), Devi Parikh (VT), Andrew Gallagher (Cornell), Larry Zitnick (MSR)

2. Edge types

3. Edge types Albedo Depth Surface normal Shadow

4. Which are useful for recognition?

5. Roberts, 1965

6. Huttenlocher and Ullman, 1987

7. Jain and Farrokhnia, 1990

8. Kanade, 1981

10. Eitz etal., 2011

11. Cole etal., 2008

12. Can we isolate the effect of different edge types?

13. Synthetic scenes 46 scenes, Google 3DWarehouse

14. Synthetic scenes Black and white

15. Synthetic scenes Albedo

16. Synthetic scenes Gray surface

17. Synthetic scenes Gray shaded

18. Synthetic scenes Depth

19. Edges RGB Albedo Depth Gray surface Gray shaded

21. Human studies 265 objects, 11 scene types, 10 subjects per test

22. Results

23. Results

24. What type of edges do our edge detectors find?

26. Precision Depth Surface normal Shadow Texture

27. Recall Depth Surface normal Shadow Texture

28. Can we further improve our edge detectors?

29. Contour detection Parikh and Zitnick, CVPR 2012

31. Conclusion Depth and surface normal edges are most informative for recognition Most automatic edge detectors find a significant number of albedo and shadow edges Better methods for evaluating edge detectors? Are lines important or just edges?