Skip to main content

Hi everyone,

I’m looking for a rapid way to view and categorize images - something that can be done with keyboard inputs, ideally. I’m not talking about annotating parts of an image, but instead being able to select from a Single Select field, and to do so quickly. This is being used for image classification for building a training dataset.

I am currently using an Interface and also customized cards (which at least allows for keyboard scrolling between records but is still clunky).

Thanks for any tips!

Not that it matters, but curious - is it a training data set for humans or AI models?

Not that it matters, but curious - is it a training data set for humans or AI models?


It’s for procuring a training set for AI models.

It’s for procuring a training set for AI models.


Cool. It’s not often Airtable is called into service in this way because of the record limitations and training sets are typically pretty large.

Cool. It’s not often Airtable is called into service in this way because of the record limitations and training sets are typically pretty large.


Really? That’s a little surprising to me, but maybe our use case is niche enough. AT seems like a natural fit for us! We have no more than 1000 images for each of our classifications (some are squarely in the hundreds), and it’s specialized enough we cannot rely on an annotation service.

Really? That’s a little surprising to me, but maybe our use case is niche enough. AT seems like a natural fit for us! We have no more than 1000 images for each of our classifications (some are squarely in the hundreds), and it’s specialized enough we cannot rely on an annotation service.


Ah, I see, then Airtable could be ideal. We train data sets that start at 50,000 images and if the record ceiling doesn’t get us, the aggregate images size will. We’re also seeing some advantages to zero-shot inferences which require no training or model tuning.

You might want to consider a script block for tagging images like this one.

Terms & ConditionsAccessibility statement