3-Sweep: Extracting Editable Objects from a Single Photo, SIGGRAPH ASIA 2013

Channel:Daniel CohenOr · Length:5:51 · Published:2013-08-13 · Views:2295220

Watch on YouTube

Summary

A SIGGRAPH Asia 2013 research demo of 3-Sweep, a tool for extracting editable 3D objects from a single photo using a few user strokes. It predates modern AI image tools.

Why HN recommends it

HN recalls it as a mind-blowing pre-AI demo of 2D-to-3D object extraction.

this single video blew my mind back in *2013* and I find myself thinking about it often. — HN, 2024

Computer vision, speech & NLP · Computer graphics & rendering

Subtitles (excerpt)

we introduce an interactive technique for modeling 3D man-made objects by extracting them from a single photograph our method intelligently combines the cognitive ability of humans with the computational accuracy of the machine we first demonstrate the use of three sweep on basic shapes to extract an object from a given photo the user uses three Strokes defining the three dimensions of the object Two Strokes Define the parts profile and the third one sweeps it along its main axis this procedure progressively defines a 3D body that snaps to the shape outline in the photo this example shows modeling a tap at after modeling we map the texture from the image and edit the parametric model to achieve image editing before manipulating the object the background is completed using the patch match algorithm we change its rotary switch to cruciform and rotate it then we duplicate the tap and rotate it and attach it to a different side of the wall this example shows modeling a telescope while modeling each part geosan relations such as parallelism concentric and orthogonality are automatically recovering after modeling we can easily edit the image part as it was in 3D here we show the rotating of the lens

On Hacker News

SIGGRAPH from over a decade ago has entered the chat... [VIDEO] It may not be AI, but this single video blew my mind back in *2013* and I find myself thinking about it often. — alfiedotwtf, 2024-10
I would love to see this combined with object extraction to make 3d models from the measurements. Here's an amazing demo: [VIDEO] — cookingrobot, 2013-09
>SIGGRAPH from over a decade ago has entered the chat... > > [VIDEO] A version of this was available in Photoshop for a long time, but I think the feature was deprecated and removed completely this year. I had used it for a few things here and there, but dedicated 3D tools were much better if you were working in that space. — leviathant, 2024-10
That is correct. Sometime in the future, consumers will have software that can turn 2D images into simulated 3D [0]. It's still in dev mode right now, though. [0] [VIDEO] — trendia, 2016-10
Lots of interesting research being done on this, keep an eye out for stuff from SIGGRAPH later this month. Here's an example from SIGGRAPH Asia: [VIDEO] . — carlosdp, 2016-07

Related videos