Cleaning algorithm finds 20% of errors in major image recognition datasets

Re:Cleaning algorithm finds 20% of errors in major image rec... by frenchie4111 on Thursday April 16, 2020 @01:50PM (#22891803)

Best I can tell, they are using the ML model to detect the errors. Isn't this a bit of an ouroboros? The model will naturally get better, because you are only correcting problems where it was right but the label was wrong.

It's not necessarily a representation of a better model, but just of a better testing set.

[ Parent | on HN ]

Re:Cleaning algorithm finds 20% of errors in major image rec... by groar on Thursday April 16, 2020 @01:51PM (#22891810)

If I understand correctly they actually did not change the test set.

[ Parent | on HN ]

Re:Cleaning algorithm finds 20% of errors in major image rec... by frenchie4111 on Thursday April 16, 2020 @01:53PM (#22891824)

Ah, I guess I missed that

[ Parent | on HN ]