If you think computers are still quite "stupid" when it comes to doing certain things we take for granted (like identifying a simple object in a photo) you're completely right. However, image processing has been evolving a lot lately... and you'll be surprised with what this 3-Sweep technique can do!
When we look at a photo we can easily identify each individual object and infer its 3D form, orientation, volume, etc. For a computer, a photo is simply a collection of pixel, RGB data that is mostly meaningless unless it can be compared to some other bunch of data.
This 3-Sweep can, with a little assistance from a human operator, recreate the 3D object shown on the photo, and easily modify it in real time. You really have to watch the video to see what it can do (and also some of the things it still can't do... for now).If you thought you couldn't believe everything you see due to photo edition programs like Photoshop, then you'll definitely know we won't be able to trust any photo, even those that seemed to difficult to fake in Photoshop - like moving a reorienting objects in 3D space inside the photo.
Can't wait for a public version of this 3-Sweep to be available so we can all give it a try. :)
Nice :D
ReplyDeleteBTW, Have you seen this: http://www.123dapp.com/catch
Works like a charm if you have enough pictures :)
Yes. There were several "multi-photo to 3D" programs. The trick is doing it with a single photo. :)
Delete