4 ms·
You highlight the big challenge with computer vision and other computerized identification of things, regardless of the underlying technology. There is no "ment
by version_five 5y ago
You highlight the big challenge with computer vision and other computerized identification of things, regardless of the underlying technology. There is no "mental model" of what makes sense against which to check observations. There can be coded rules, but if something passes them, there is no extra sanity check like a person would do. Put another way, models have no common sense, and we don't really have a meaningful way to give it to them. This is going to remain a problem for self driving cars and is a big part of the reason why they have not lived up to expectations.
- aabhay 5y agoHumans use stereo vision to overcome things like this. I think some kind of stereoscopic strategy would be a feasible strategy, no? Why revert to the age-old argument of "common sense"?
- patch_cable 5y agoI have one eye and have no problem not making these kinds of mistakes.
- nyx 5y agoHumans have more than just stereo vision sensors. Like GP said, we also have robust schemata, seeded by biology and improved over our entire lives, that allow us to make sense of the things we see. These heuristics are trivial for us to access, but relatively difficult to describe and implement in a software system. That's what "common sense" means in this context.
- clon 5y agoStereo vision is overrated. One of my buddies is a pilot with 1 eye (since birth) and around 11 thousand landings. Top of his game. There is more to depth perception than stereo vision for sure
- jazzyjackson 5y agoWe have a lens that has to flex to bring far away objects into focus, so I figure that can give us some feedback.
- wizzwizz4 5y agoFor sufficiently distant objects, this is completely irrelevant. This lets us roughly gauge the reciprocal of distance.