I really love the combination of the ‘Ken Burns’ effect to bring images to life, with a way to cut the image up into layers, and then move each layer separately.
It’s usually a lot of work in photoshop.
But now, with the amazing work from Simon Niklaus et al, it seems that ML models can perform this task by using segmentation to figure out foreground from background and inpainting to fill in the missing pixels.
It’s ingenious.
I played with the image from above, and got this:
Still a bit of work to be done, and I’m hoping I can tweak for higher resolution in the video, but the effect is definitely stunning…considering it’s entirely automated!