Replying to @mhmazur
it's a gpu implementation for morpho programs, using bounded per-cell info and tree-encoded buses
paradigms-of-intelligence.gi…
2
26
Replying to @rgilman33
This reminded me of a blog post by @EvMill, it seems to be a good way to solve this problem
evanmiller.org/attention-is-…
Replying to @jperldev @NVIDIAAIDev
Thanks! It's from a quick phone video, depth of field is simulated afterward.
I didn't notice the cicada molt until re-watching just now, there may have been some very slight movement in the flowers preventing a fully detailed reconstruction.
1
Trying out the depth of field effect, it works much faster than I expected it to
@NVIDIAAIDev #InstantNeRFSweepstakes #volumetric #neuralrendering
Quick scan of a canoe/kayak ramp I saw today
@NVIDIAAIDev #InstantNeRFSweepstakes #volumetric #neuralrendering
Replying to @jonstephens85 @NVIDIAAIDev
Yeah, I just noticed this in another render. Using 25fps should avoid it but a more general fix would be nice to have.
I also noticed that the first frame (which you see in embeds) is very blurry. That can be fixed by adding a call to testbed.render before the loop.
Replying to @bilawalsidhu @NVIDIAAIDev
Yeah, I want to try capturing some tiny plants in high detail next time
First render with InstantNeRF, a lion garden statue
@NVIDIAAIDev #InstantNeRFSweepstakes #volumetric #neuralrendering
I made a simple Pix2Pix animation program with #p5js and #TensorFlowJS
Link and source code here:
parameterized.github.io/proj…
Replying to @ompuco
Have you seen this?
nvlabs.github.io/wc-vid2vid/
It probably wouldn't run in real-time but it's related to what you're doing
1
Replying to @parameterized @zachlieberman
If you want the output in the range [-1, 1], sin(pi/2 * x) would work better



