I'm trying to evaluate human accuracy on ImageNet test set. Really tricky to make interface to label among 1000 classes... #kindastuck
Any ideas? @arfon @orbitingfrog MT @karpathy Trying to evaluate human accuracy on ImageNet... tricky to make interface to label 1000 classes
![[𝚘𝚋𝚓𝚎𝚌𝚝 𝙾𝚋𝚓𝚎𝚌𝚝]](https://pbs.twimg.com/profile_images/835042549156855809/fky6YRY3_normal.jpg)
-
-
@michael_nielsen@arfon@orbitingfrog So far building nice web interface, with hierarchical breakdown following imagenet hierarchy. bleh. -
@karpathy@arfon@orbitingfrog Seems like the kind of problem@the_zooniverse people might have tackled previously@chrislintott
End of conversation
New conversation -
-
-
@orbitingfrog@karpathy@chrislintott Figuring out human-level perf on important ML tasks sounds like s'thing@the_zooniverse wld be good at -
@orbitingfrog@karpathy@chrislintott Definitely true for ImageNet: personally I misclassify many images (it's tough), so don't know what
End of conversation
New conversation -
-
-
@michael_nielsen@arfon@orbitingfrog@karpathy another idea: free text input producing a list of classes related via WordNet, etc.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.