3 ms·
I'm usually a big critic of AI, and I remember seeing people complaining years ago about increasing framerates with AI[0], but maybe this could be a good usecas
by internetter 3y ago
I'm usually a big critic of AI, and I remember seeing people complaining years ago about increasing framerates with AI[0], but maybe this could be a good usecase. You do, say, half the cameras, and then some crafty solution to interpolate (filling in the edges with AI, some sort of angle adjust)
[0]: https://www.youtube.com/watch?v=_KRb_qV9P4g https://www.youtube.com/watch?v=_KRb_qV9P4g
- estebank 3y agoThe project already used frame interpolation (through DaVinci Resolve) to double the number of frames in the video. Plus a low-tech "inside the wall" panned and blurred frame. Given the set-up, fancier scene reconstruction to build a 3D model of the environment could be combined with texture reprojection to improve the interpolation and allow the "inside the wall" frame to be perspective corrected, but that would be much more involved on the software side than what was done here (and beyond what I could do too).
- zoogeny 3y agoThe tech to look out for in this space is called NeRF (Neural Radiance Fields). Just do a quick search on that term and you will find a whole bunch of videos that reconstruct 3d scenes using digital stills. I just saw a paper a couple of days ago showing advances in reconstructing human expressions that change over time using advanced techniques.
- riotnrrd 3y agoYeah, the Frugal Man's Bullet-time is a DSLR, Instant NGP from NVIDIA, and COLMAP.
- 7373737373 3y agoI want to show the creator this specific video on DAIN (Depth-Aware Video Frame Interpolation): https://youtu.be/sFN9dzw0qH8 https://youtu.be/sFN9dzw0qH8 ! https://github.com/baowenbo/DAIN https://github.com/baowenbo/DAIN