“Dogs & cats morphing together—mass hysteria!”

It’s that thing where you wake up, see some exciting research, tab over to Slack to share it with your team—and then notice that the work is from your teammates. 😝

Check out StyleAlign from my teammate Eli Shechtman & collaborators. Among other things, they’ve discovered interesting, useful correspondences in ML models for very different kinds of objects:

We find that the child model’s latent spaces are semantically aligned with those of the parent, inheriting incredibly rich semantics, even for distant data domains such as human faces and churches. Second, equipped with this better understanding, we leverage aligned models to solve a diverse set of tasks. In addition to image translation, we demonstrate fully automatic cross-domain image morphing

Here’s a little taste of what it enables:

And to save you the trouble of looking up the afore-referenced Ghostbusters line, here ya go. 👻

Leave a Reply

Your email address will not be published. Required fields are marked *