Can Edge Impulse be used as an accelerator for models which are to be deployed on a server or PC?
The shortcuts the Edge Impulse provides and its correspondence with Keras models would seem to make it appealing for developing more complex models and deploying them to a PC in Linux or Windows.
Would this be a misuse of the technology? Also can RNNs or other model types than CNNs be practical with Edge Impulse?
Hello @IML2,
Can Edge Impulse be used as an accelerator for models which are to be deployed on a server or PC?
Yes, we also support deploying models on linux x86, we will add that to the documentation soon. You can download the .eim model from the studio or directly from the Edge Impulse Linux CLI and then integrate your model using one of our SDKs (https://docs.edgeimpulse.com/docs/edge-impulse-for-linux#sdks)
Would this be a misuse of the technology?
Not sure but we focus on Edge ML.
Also can RNNs or other model types than CNNs be practical with Edge Impulse?
Not at the moment but I remember @janjongboom saying that it could work on Linux. I don’t think anyone tested it though so it’s definetely not guarantee.
Regards,
Louis
Hi @IML2,
All deployed models from the Edge Impulse Studio can be used on a Server or PC, check out our documentation on running your impulse locally:
- Edge Impulse for Linux: https://docs.edgeimpulse.com/docs/edge-impulse-for-linux
- Running your impulse locally on your desktop computer: https://docs.edgeimpulse.com/docs/running-your-impulse-locally
- Through WebAssembly in your browser: https://docs.edgeimpulse.com/docs/through-webassembly