Re:Re-creating Disney's sodium vapor process featuring Paul ...
by jmpman on Monday April 08, 2024 @08:26PM
(#39975130)
I’ve been doing something similar using “cheap” Chinese band notch filters. Had no idea about this Disney process. Also using this for AI training purposes. It’s amazing what you can buy on Alibaba, fully custom frequencies.
[ Parent
| on HN ]
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by krasin on Monday April 08, 2024 @09:06PM
(#39975357)
I would be interested in some links!
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by sixothree on Tuesday April 09, 2024 @12:48AM
(#39976250)
After you using sodium vapor out a different light source?
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by jmpman on Tuesday April 09, 2024 @11:19AM
(#39980384)
Different frequency.
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by p_noumenon on Wednesday April 10, 2024 @10:25AM
(#39991054)
Any thoughts on whether some frequencies are superior to others for various reasons, or what strengths/weaknesses different frequencies have?
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by jmpman on Thursday April 11, 2024 @09:33AM
(#40001963)
The frequencies I’m using are very specific to the subject I’m filming. It’s closer to a machine vision problem, and it isolates the desired object almost perfectly from the background. I’ve considered filing for a patent, it works that well.
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by fnands on Tuesday April 09, 2024 @05:07AM
(#39977577)
> Also using this for AI training purposes.
Cool! Training for what? You can't just drop a juicy comment like this and then leave us hanging ;-)
Re:Re-creating Disney's sodium vapor process featuring Paul ...
by speps on Tuesday April 09, 2024 @06:37PM
(#39985079)
It's explained in the video. Keying algorithms are getting better thanks to AI image techniques. However the training data they use is based on green screen technique so has the same issues at times. Using the sodium vapor method to film training data that looks correct to start with would improve these AI algorithms.