RZA2M food-recognition-demo

Hi All.

https://github.com/renesas-rz/rza2m-food-recognition-demo

I use it according to this manual.But when I burn to GR-MANGO, this error will appear first.

After some time, another error is displayed.

The first contact is related to AI, where is the problem?

Thanks!

Parents
  • Hi Nazukolin,

    the AI part of the demo demo is based on TinyML libraries plus a dedicated DRP library for convolutional layer.

    TinyML starts an 'interpreter' that handles the inference. At the beginning, the interpreter needs to be initialized (invoked) and this is what is failing on your side.

    I guess you exchanged the neural network. If that's the case, please carefully check the interpreter setup. That's documented as well, but the combination with the DRP library is a bit tricky.

    The first Error indicates that the required TFLite library for a dedicated layer is not available (it may be that the library is available, but the configuration does not fit).

    The 2nd Error is related to the interpreter size definition that seems to be 204, but the neural network is bigger and would require 1000.


    Did you check the binary of the demo ?
    That should work from scratch.

Reply
  • Hi Nazukolin,

    the AI part of the demo demo is based on TinyML libraries plus a dedicated DRP library for convolutional layer.

    TinyML starts an 'interpreter' that handles the inference. At the beginning, the interpreter needs to be initialized (invoked) and this is what is failing on your side.

    I guess you exchanged the neural network. If that's the case, please carefully check the interpreter setup. That's documented as well, but the combination with the DRP library is a bit tricky.

    The first Error indicates that the required TFLite library for a dedicated layer is not available (it may be that the library is available, but the configuration does not fit).

    The 2nd Error is related to the interpreter size definition that seems to be 204, but the neural network is bigger and would require 1000.


    Did you check the binary of the demo ?
    That should work from scratch.

Children
No Data