Ten Questions on Computer Vision from VALSE

In the 2nd Vision And Learning Seminar (VALSE) held by Chinese Academy of Sciences, participants raised ten questions on Computer Vision after free discussion. I translated and summarized them as follows:

1. Does Computer Vision have real solid theoretical foundation? Except from Marr's non-perfect visual computing theory, what kind of theoretical foundation do we need to set up now?

2. What is the ultimate questions in the Computer Vision field, which have to be high-level enough? Let's look at some ultimate questions in Computer Science field for reference: What is computable? P=NP? What is intelligence? What is information? (how) can we build complex system simply?

3. In Computer Vision and Machine Learning field, math and data, who is the king? Is big data the ultimate solution to Computer Vision problems?

4. Is biologically inspired models the hope for computer vision community?

5. What are the cans and can'ts of sparse representation? How far can it go? What else can be done and what is the killer application?

6. How far can local feature go in visual recognition problem?

7. Ridiculous Computer Vision reviewers? Lecun's anger in CVPR 2012 rebuttal: what happens to the evaluation system in computer vision community and how can we improve it?

8. The challenge from industry: Is the best technology in industry? Does the occupation of specific resources, especially big data and big platforms in industry form an invisible academic barrier?

9. What is the biggest progress in the Computer Vision and Machine Learning community in the past ten years? What about the top 5 progress? And the top 10?

10. Comparing with the researchers in Europe and America, what is the gap and hope of Chinese Computer Vision community?