is computer vision like common sense?

[gusl]

Why No Vision?, from Push Singh's blog (he is a student of Marvin Minsky)

Why is it that computer vision has proven to be such a difficult problem? The strange thing is that computer graphics, which one might regard as the inverse problem, is rapidly closing in on achieving photorealistic rendering of scenes. I'm also puzzled because recognition problems are typically simpler than generation problems. It's certainly true that computer graphics has benefited from much commercial development and Moore's law, but faster computers should help recognition tasks as well.

One idea is that vision suffers from the same kind of problem as does commonsense reasoning, namely, the lack of large scale knowledge bases about the kinds of objects and materials in the world, what they look like from different angles and under different lighting, and so forth. But if this is the case, and computer graphics has advanced so far, it should not be difficult to generate a suitable such corpus with a moderate investment -- a corpus of images, ground truths in terms of 3d and other types of surface models, and connections to more general commonsense knowledge.

Comments

[altamira16.livejournal.com]

I think that it may be the fact that it requires some level of cognition. It's why speech recognition is okay, but speech understanding is not.

Basically, you could probably split vision into problems of features such as edges, textures, and colors, but how do understand the composite of all these features?

These problems are solved by black box machine learning, state vector machines, type systems that do arrive at the correct results but the feature sets they rely on may be totally contrary to what people use to make the same type of decisions.

I believe I remember someone solving a facial recognition problem using a hidden markov model and covering features of the face and finding recognition rates after certain features had been removed. It turned out that hairline was huge for the machine to determine the difference between male and female faces. But typically, that is not what we would think of as a feature to help us differentiate men and women.

There are limited solutions in terms of OCR, but the nature of the recognition and generation problems are very different.