How we can improve processing in real time for video?

yonomitt / Anime-fy-Yourself

Code and documentation for my talk Anime-fy Yourself Using Core ML

MIT License

4 stars 0 forks source link

How we can improve processing in real time for video? #4

Open ioskevinshah opened 2 years ago

yonomitt commented 2 years ago

Good question! I’ve been thinking about this and just opened issue #5.

Another thing that can be done is to use the new performance tab in Xcode 14 beta, to see which layers of the model are running on which devices (CPU, GPU, ANE). Those results would inform what layers could be optimized (if any).

ioskevinshah commented 2 years ago

I already did, but I don't how to improve it in python script.

yonomitt commented 2 years ago

I don't have any devices running macOS beta or iOS beta, so I haven't been able to run the performance analysis, yet. Once I get some device on beta, I can try and see if anything can be improved.

In the meantime, trying the camera improvement from issue #5 should be easy.

ioskevinshah commented 2 years ago

Thanks