Google-Landmarks: A New Dataset and Challenge for Landmark Recognition

Looks like we here in picturesque New Zealand may have more landmarks than Africa and the Middle East combined!

More seriously, I didn't see anything about the providence or license of the pictures, or how they ended up in this collection. The Kaggle competition data doesn't seem to contain the actual photos, just links.

The photos might be from contributions on Google maps, which users grant rights to Google.

When will they release their recaptcha datasets, and their voice datasets? I'm really happy to see this, but I feel like we're being tossed crumbs here.

I wonder if it is data from Ingress?

If there are trash bins marked as landmarks then it's Ingress.