Unsupervised learning of probably symmetric deformable 3D objects from images • Hacker News Board

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#21

Post by vidarh » Wed, Mar 25, 2020, 4:26 PM UTC

st_goliath wrote: post #9

jcims wrote: post #3

Stuff like this can obviously be used to make things like deepfakes 'better'. But i think it might be cool for creating virtual meeting rooms, where you can take a webcam shot of a persons face, normalize the skin tones for lighting, map to a 3d surface, then relight it for the virtual room. When you can rig the meshes to drive each other you could wear 'masks' of other peoples faces (or critters).

> But i think it might be cool for creating virtual meeting rooms, where you can take a webcam shot of a persons face, normalize the skin tones for lighting, map to a 3d surface, then relight it for the virtual room. You might be interested in project HeadOn at TU München: https://www.niessnerlab.org/projects/thies2018headon.html Justus Thies gave a presentation at our University about a year ago. IIRC they don't use…

One of Vernor Vinge's novels describes 3d models transmitted as part of video conferencing, and re-skinning used to fool an adversary (while depending on causing a noisy, reduced bandwidth connection to make it plausible why the model is imperfect, I seem to remember).

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#22

Post by Frost1x » Wed, Mar 25, 2020, 4:28 PM UTC

st_goliath wrote: post #9

jcims wrote: post #3

Stuff like this can obviously be used to make things like deepfakes 'better'. But i think it might be cool for creating virtual meeting rooms, where you can take a webcam shot of a persons face, normalize the skin tones for lighting, map to a 3d surface, then relight it for the virtual room. When you can rig the meshes to drive each other you could wear 'masks' of other peoples faces (or critters).

> But i think it might be cool for creating virtual meeting rooms, where you can take a webcam shot of a persons face, normalize the skin tones for lighting, map to a 3d surface, then relight it for the virtual room. You might be interested in project HeadOn at TU München: https://www.niessnerlab.org/projects/thies2018headon.html Justus Thies gave a presentation at our University about a year ago. IIRC they don't use…

These photogrametric, structure from motion, structured light, etc. techniques have been around awhile so I don't think this changes too much though it may make it a bit easier to generate realistic depth maps for other purposes.

At some point you can probably reconstruct large portions of scenes in existing movies and change perspectives, especially if you had good techniques (perhaps AI based) for filling occlusions in the data.

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#23

Post by hirako2000 » Wed, Mar 25, 2020, 4:36 PM UTC

coding123 wrote: post #5

I think we slammed it dead.

If it was truly processing on the browser, we wouldn't have jagged it so easily.

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#24

Post by usrusr » Wed, Mar 25, 2020, 4:40 PM UTC

thomasahle wrote: post #8

Video here: https://youtu.be/5rPJyrU-WE4 Can anyone explain why people would use text to speech for something like this, when they have perfectly good voices themselves?

In addition to the possible reasons stated by peer comments, revisability. One-click builds are just as attractive for writing as they are for programming.

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#25

Post by nalaka » Wed, Mar 25, 2020, 4:40 PM UTC

Looks like HN has broken their server. I am getting a "Failed to send request to server!" error when I upload an image.

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#26

Post by nalaka » Wed, Mar 25, 2020, 4:42 PM UTC

thedance wrote: post #7

ijpsud wrote: post #2

Not loading properly for me :( I think server may be being overloaded by HN traffic?

Unsupervised Static Asset Server Converts Probably Simple HTTP Requests to 404 Errors.

Yes- you nailed it. :)

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#27

Post by dang » Wed, Mar 25, 2020, 4:46 PM UTC

h91wka wrote: post #11

Please fix the title of submission. It is " Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild". No anime examples to be found in the paper :(

Submitted title was "Deep network in the browser converts to 3D any image of a person/cat/anime face". Changed now.

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#28

Post by BorisTheBrave » Wed, Mar 25, 2020, 4:47 PM UTC

thomasahle wrote: post #8

Video here: https://youtu.be/5rPJyrU-WE4 Can anyone explain why people would use text to speech for something like this, when they have perfectly good voices themselves?

I've found it's surprisingly hard to get a satisfactory recording setup - noise, volume, echoes.

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#29

Post by Karuma » Wed, Mar 25, 2020, 4:56 PM UTC

deckar01 wrote: post #10

"We store a copy of the uploaded image ..." Is this really happening with a "deep network in the browser"? It looks like it is happening on a server, then the 3D result is viewed in a browser.

Yeah, I'm glad they changed the completely inaccurate title. Right now it doesn't even work due to server overload...

Re: Unsupervised learning of probably symmetric deformable 3D objects from images

#30

Post by jotaf » Wed, Mar 25, 2020, 5:05 PM UTC

nalaka wrote: post #26

thedance wrote: post #7

Earlier quoted context omitted.

Unsupervised Static Asset Server Converts Probably Simple HTTP Requests to 404 Errors.

Yes- you nailed it. :)

Should've updated the title! Now it converts HTTP requests to proper 3D meshes (seems to work again).