Many companies naturally generate image data in the same way they naturally generate tabular data, but historically it's been much harder to make productive use of image data. In the past, making sense of image data usually required some kind of deep learning pipeline. If you wanted, say, to know how many blue shirts you have in a product image catalogue, somebody would have to fine tune an image classifier and set up a Python pipeline specifically for that query.
With Coactive, an analyst could simply write "select count(*) where has_blue_shirt = 1" after doing a couple hand-labelings. Importantly, the analyst isn't dependent on an engineering team to do this. We're excited to see Coactive make image data much more accessible for analytics teams.