Hi all,
I have a question rearding the deployment of any model from Edge Impulse.
If my model occupies around 16KByte of flash in some Arm Cortex processor and the data acquired are about 50MByte. The deployed model as an Arduino Library or C/C++ one would have the model only or the model and data acquired also??
Best,
Nawaf
Hi @EngN7,
You do not need to store the data with your model once you deploy. Remember that the model attempts to generalize meaning from the data so that you do not need to keep track of the raw data.
When you deploy, you would need the flash space and RAM for storing and processing the model (16 kB for flash in your case). You also need to have enough RAM for storing a single sample prior to processing.
For example, if you record a 1 second clip of audio for keyword spotting, you would need the RAM and flash for the model as well as that clip (e.g. at 8 kHz sample rate and 8-bit depth, you’d need 8 kB of RAM available for the sample clip on top of the model).
Thank you, Shawn, for your quick response!
So, this means that the deployment is only for the model information and the data acquired is only for the training purpose within the webiste of Edge Impulse and not stored in the hardware whatsoever.
That is correct–the data used for training is not part of the deployment process.