Many good ideas will not work well on MNIST (e.g. batch norm). Inversely many bad ideas may work on MNIST and no transfer to real CV.
Lots of issues with MNIST, but most of all, it is really not representative of CV tasks. Please at least try CIFAR10, of comparable size.https://twitter.com/goodfellow_ian/status/852591106655043584 …
-
-
-
Besides, MNIST cannot serve as a benchmark to compare different ideas. But I suspect that's why people love it so much. Competing is hard.
- Show replies
New conversation -
-
-
MNIST is Hello World.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Disagree. The training time of a cnn in mnist has nothing to do with cifar10. Most often the intelligent thing is to implement mnist first.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
People still benchmark k-NN search on MNIST, which is even funnier, b/c any reasonable bruteforce search is 2-10 ms, single core.
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.