It is impressive what can be done with this technology, yet this would only be at the very surface of artificial vision, IMHO*, because:

• It is not completely automatic, the user has to specify some constraints to be kept
• Reconstructing a guy (as an example) is great, but still:
  • He should be recognized as an object
  • Patterns should be found among different objects to find out that people are people and at the same time different persons, while other things are not people.
  • Patterns should be found in the same person, as he may change due to movement, costumes, etc.
• The video runs at 5x for some time, and although it is not required to paint anything after seeing it I’m assuming the painting part is only a small fraction of the algorithm. Still more power is needed.

So artificial vision is not finished yet (no surprise), I hope this video may help to understand how much computational power may be required for artificial vision and the great job people are doing in that field with the available power, I’ll try to post more videos for this matter.

* IMHO here means in my really really humble opinion.

PD: There is another video that I failed to notice that was different, this “new” video has some additional technical details.