Many years ago (before the advent of OpenStreetCam or Mapillary) I proposed a federated scheme where various entities could obtain their own dirt-cheap http hosting (backblaze?) and simply submit indexes (or make available in an easily discoverable fashion) of the image metadata for centralized tools to pull together, but this was dismissed as lunacy.
I still don't think it would be too tricky to do and would nicely decouple the fundamental storage/hosting problem from all the other problems that might need solving.
What I thought was particularly interesting was how EXIF data tends to be near the beginning of a JFIF file which could allow "cheap" bulk scanning of jpeg metadata by using http 1.1 ranges.... who knows.
I think it will be a while before there is a community project accepting many gigabytes of photos every day.