Converting high res photos into 3D models fully automatically

Hi,
use common sense.
If there existed fully automated process "2D photo(s) to high-quality 3D model" the market would be completely flooded by thousands or milions of 1$ models.
Yes there is 3D scanning (photogrammetry) but this is not suitable for all types of 3D models - you will never get accurate model of car by 3D models and by semi-automated retopology tools.

Any technology (at this moment) cannot fully replace artists work - speaking about accurate models - it can only help to the artist.

It depends how accurate model needs to be, what it will be used for etc. Judging by your post, you need 3d presentation of products that you are going to sell on you website (correct me if im wrong). So here is the thing, there are some softwares that can generate 3d model from photoS not photo (alghorythnm that can generate depth from using a single photo doesn't exist) but end result usually need a lot of work in order to get that 3d model to a usable level, most of the time it's easier and less time consuming to model that object by hand than to fix it's topology that is generated by software, and that work is done by human factor. Keep in mind performance of the website because 3d models will have big impact on its performance if they aren't done properly. First software that comes to mind is Autodesk 123d, i'll provide the link of what it is at the end of the post, it's self explanatory. The closes thing that i saw to something that can generate almost clean topolgy 3d model is a system made by Team Bondi (game company), system use 32 hi-res cameras specificaly aranged with custom processing unit and can capture anything , it can also capture movments so its used as MOCAP for animating purposes. This "toy" cost multiple milions of dollars and even this is far from perfect.

https://www.youtube.com/watch?v=NsBg-m2hrIM

I certainly hope that software doesn't exist, else I'd be out of a few bucks. I can't see it ever existing, since any model needs 3d information. You can probably get software to interpolate a photo of a person's face and make up the missing data of the back of the head, since MOST heads are round. (lol)
It still would be hit and miss on any type of accuracy.

TazMan2000
never say that it can't happen, because the possibility is definitively there.
Even although they are still not usable for making good 3d models, computer programs that involve neural networks are already being invented with the ability to make up the "missing information" you mentioned.

Basically, such a program first runs recognition software to know what the image is about, then it uses internet to find thousands of pictures of the topic, then it filters out relevent images of the right angles.
Then using the information from hundreds of selected photo's, it could compose a 3d model without it missing out on any information.

As far as i know, there are no programs like this for 3d objects "yet" but i have seen some very impressive results when it comes to 2d images/photo's that have been produced by a computer from just a text entry alone.

When the first good version for 3d models will be developed? i don't know, might be 10 years, might be a 100. But the possibility of it being developed is very real.

Polymath-Maniac,

You're right. A version of that sort of tech exists today as photogrammetry. I'm sure it wouldn't be too hard to program some AI into software that can create sharp angles for some pieces and softer ones for something more organic.
I was just saying that it can't be done 'well' from just one photo.
I've seen some of the design results from these supercomputers, and indeed they are impressive, however, I'm thinking the results we are seeing are picked out of several hundred failures.

Maybe, but i think it cost so much and not effective application