3 ms·
PS: The human being beaten was a guy from Stanford who did the validation set for this specific image recognition task a while ago (he somewhere said that he co
by singularity2001 12y ago
PS: The human being beaten was a guy from Stanford who did the validation set for this specific image recognition task a while ago (he somewhere said that he could reduce the error rate to 3% given a second chance ;)
This same week Microsoft released a similar result, but without such an important new approach for networks.
The important chart really is on page number seven.
- karpathy 12y agoYou're referring to my google+ post from yesterday :) https://plus.google.com/+AndrejKarpathy/posts/dwDNcBuWTWf https://plus.google.com/+AndrejKarpathy/posts/dwDNcBuWTWf The short story is that I tried for 500 images and then got 5.1% error on the test set. Looking at my mistakes, I then tried to differentiate two sources of error: a genuine problem with the dataset (e.g. many correct answers in an image, or incorrect label), and errors I felt could be eliminated by an ensemble of very committed humans who were even better than me at classifying dogs :) And that optimistic error rate is approx 3%.