great. i agree that this potentially will be the coolest thing:
ok. i agree that if you’re primarily looking for a training set, then you’re probably just going to download the full set of metadata files or do something with an AWS RDS. i was originally thinking that for other potential uses, it might be more useful to start by getting observation data via the API or observation export, rather than starting with the observation metadata file. i was thinking there might be some middle area between the 10,000 observation limit of the API and the 200,000 limit of the observation export, where it might be nice to pair the photo metadata file with the results of an export, and folks most likely to employ that kind of workflow might already have a particular photo license or set of licenses in mind. but upon reflection, i suppose if you’re going to download and use >10,000 photos, then downloading and using a large photo metadata file probably isn’t going to phase you.