Back to DrivenData | Blog

Confusion about "healthy" and "irregular" metal

Hey everyone,

while browsing the training images I got confused about the definition of “irregular metal” and “healthy metal”, especially regarding the images in the training set for Dennery. I wonder why this is labelled as “healthy metal” (id=7a328e18):
7a328e18

On the other hand, some healthy-looking roofs end up as “irregular”, although they look pretty clean, e.g. [id=7a1f4e16].

For both images, labels are marked as verified. Especially in the Dennery dataset there are a lot more examples for this seeming inconsistency. Is this really an issue with wrong labeling or did I just get the definitions of healthy and irregular labels wrong?

Any clarification would be appreciated :slight_smile:

2 Likes

Hi @jruether looks like we both are on the same boat with respect to confusion in dataset if u have got better clarifications with respect to defining labels plz do share in the forum. I hope it does not come under solution sharing.