Memory Error, out of fast Frame Buffer Stack - H7

i’m going to run machine learning example ’ TensorFlow Lite Mobilenet V1 Example ’ on OpenMV H7
and then it show up an error ‘MemoryError: Out of fast Frame Buffer Stack Memory! Please reduce the resolution of the image you are running this algorithm on to bypass this issue!’
but i’m not edit anything on this code?

what should i do?
thank you

Hi, the mobilenet example loads the network to memory and only works on the H7-Plus (with external SDRAM).

It doesn’t even work on the H7 Plus anymore. Google broke Mobilenet support with their recent updates to TensorFlow. We are waiting for a fix.

thank you, i’m missing to read the comment on top of the example code

omg… really?
what should i do if i want to train a model, convert to .tflite and deploy on OpenMV.
now i used this model mobilenet_v1 | Kaggle
and it error like this ’ OSError: Didn’t find op for builtin opcode ‘MUL’ version ‘2’ ’

if i can use the model from tfhub can you suggest me for a model that i can re-train for my dataset?

thank you very much

Hi, you need to install the latest firmware to run that model. Which… might not have been released yet. If you can compile the firmware for the H7 and install it you should be good to go.

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/micro/kernels/all_ops_resolver.cc#L21

We added support for MUL versions 1 to 3.

i got an errro when i try to run make.py for build tensorflow-lib
‘’’ tensorflow/lite/micro/kernels/all_ops_resolver.cc: In constructor '‘tflite::ops::micro::AllOpsResolver::AllOpsResolver()’:tensorflow/lite/micro/kernels/all_ops_resolver.cc::68::35: error: ‘Register_SUB’ was not declaerd in this scope
AddBuiltIn(BuiltinOperator_SUB, Register_SUB(), 1, 2);
tensorflow/lite/micro/kernels/all_ops_resolver.cc::68::35: note: suggested alternative: ‘Register_SIN’
AddBuiltin(BuiltinOperator_SUB, Register_SUB(), 1, 2);

what wrong? i don’t know how to solve it?
and… if i can’t to do that, when the next release will come?

thank you

Did you clone the repo correctly? If you pulled the latest tensorflow then I have no idea. Things break constantly upstream. You should just clone the repo and init submodules and then it should build okay.

Also, try compiling more than once. TensorFlow fails to build the first time sometimes.