How to classify socks using a Raspberry Pi, Edge Impulse, and balena

Project help
43

Hello @stratosb Today Edge Impulse pushed a PR on their repository and now it should work. Could you please check on 32 or 64 bits?

Keep us updated :slight_smile:

45

Ok, I will.
Thank you.

47

Hi @mpous,
I deployed again the EdgeImpulse ML Model on raspberry pi 3 with balena and works correctly.
It displays the three states I have specified: pair, unpair and unknown
Thank you

1 Like
48

@mpous
Regarding the balenaOS version, the dashboard displays only the following three versions when adding a new device:
v2.58.3+rev1
v2.56.0+rev2
v2.51.1+rev1

So I don’t know which version is the 32-bit or 64-bit balenaOS.
For this project I chose the recommended version, which is v2.58.3+rev1

49

@mpous
If I wish to retrain the learning model with more pairs of socks, how can I deploy the new EdgeImpulse ML Model on the raspberry pi?
What I do now it to delete the device from the balena dashboard, download the new balenaOS image and flash it again to my SD card.
Is there a simpler and faster way to do it?
Thank you

51

@stratosb When you select the device architecture (e.g. Raspberry Pi 3 (32 bits)) then you select the compatible balenaOS versions and supervisor versions. You can see the recommended versions here https://balena-io.github.io/balenaosversions/

1 Like
52

@stratosb as far as i know, you will need to reboot your device after retraining the model. Then automatically will download the latest version of the model to the device. Is it not working for you?

55

I will try it, thank you