How is frame rate ?

[GarrickLin]

Did you only use the cpu of pi3 to run MTCNN ?

[solderzzc]

No GPU based library available on Raspberry Pi 3 for now, so it's CPU only.
One frame would cost 4-8s.

[aureosun]

您好，请问在树莓派上边运行速度怎么提高？我目前遇到了这方面的问题，想请教一下您？

[solderzzc]

@aureosun If you are using Tensorflow, you can try to quantize the module. A how to

[arasharchor]

@solderzzc
4-8s based on which image resolution?
is it possible to use GPU of Raspberry Pi?
I looked into Model Quantization, but couldn't manage to implement it on this current code.