deleted user
Unregistered
Thanks:
Given thank(s) in post(s)
Yeah deep learning would have been my only guess. Most software I read about is only half-automated and needs manual cleanup and parameters etc. (not to mention it being incredibly pricey)
Posts: 2,051
Threads: 56
Joined: 2016 Dec
Thanks: 162
Given 1011 thank(s) in 614 post(s)
Generally the ideal way to convert 2D to 3D is to generate a 'Depth Map' which is then used by software to reposition a 2D image/video file into Left Eye/Right Eye Stereo 3D. The gaps left by the conversion are then filled in.
Different 3D houses have different methods to create the depth map. StudioD uses rotoscopes while Legend3D uses a software tool originally designed for colorising B&W movies. Even the more automated processes require tons of man hours in tweaking/adjustments alongside the inpainting.
I had a little look awhile ago at this. There is software out there that will create the 3D from 2D + Depth Map, there are various experimental programs/scripts that can help to generate the depth maps. It's a case of lots of learning and huge amounts of time though.
A lot of media players and displays can do on-the-fly 2D->3D but it's usually horrible. Eyeball-hurting horrible. I think the biggest reason there is no 'consumer' version of the software available is that it will almost 100% certainly be used by people to convert movies, i.e. copyrighted and encrypted media.