A great software is eating the world story, from a family member in a traditional industry, who is not themselves a technologist but has worked for the Internet for a few years:
-
-
would it be accurate to say that some large percentage of the photos satisfy basically no properties at all (ie just blurry noise) and that excluding all of these would reduce the sample considerably?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
(but right now i would guess you’re not quite right on the scaling -you could come up with an algorithm that does decent job but it’s going to misclassify some and fail to classify others-depending on the penalty for doing that you could end up having to have humans take a pass)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
also a million images isn’t much. if you could put 100 on a big screen at the same time, how long would it take a person to click on the ones that have this feature? under a minute? now multiply this by 10,000. well under 1000 human hours
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
If a human six year old can trivially tell the difference, it is very likely that so can a neutral network, and someone with a few days investment in learning http://fast.ai can train it.
-
I agree and this is where the software eating the world story kicks in; once trained, the actual effort in doing the work is constant or scales up marginally. The cost and addressable scope for training/modeling is always improving, unlike for humans; we're stagnating.
End of conversation
New conversation -
-
-
Garbage photo image classifier. Interesting idea.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.