EdgeImpulse on RPI Linux: 32 vs 64 bit?

[ramfish]

We are running EI for Gesture and voice recognition on RPI Zero 2.

What are people insights regarding using the 32bit vs 64bit rpi os versions?

[jorduino]

Hi @ramfish ,

Interesting topic!

The inferencing should run faster on 64bit, personally not tested this yet though, let me check with the team to see what others experience was.

Best

Eoin

[MMarcial]

@jorduino Are you saying the Edge Impulse inferencing should run faster on 64-bit RPI because the deployed Edge Impulse library has 32-bit and 64-bit compiler directives within the code and as such can take advantage of 64-bit?

[Eoin]

Hi @MMarcial

The method was described in this post, adding a link to save duplication, and we can continue there if needed

ENOENT when running .eim in docker on Raspberry Pi 4 Model B Rev 1.2

@inteladata, uname will give the architecture of the host system even if running it inside a 32-bits container. You should be able to use the aarch64 binary and it will give you better performances. Just make sure to use 64-bits versions for your docker engine and container. Aurelien

Best

Eoin