ive trained and saved a keras model for detecting some specific letters and i checked if the op i used are supported on tflite.i checked and my model is working properly but every time i try to do sth with it with my Openmv Cam (H7) when the program reaches a line who has sth to do with my tflite file (for example tf.load or tf.classify) the program just stops! with no errors but the LED blinks in green for a few times and afterwards it blink white.whats wrong ? BTW im using Openmv Ide v.2.5.1
That’s a hard crash happening. The TensorFlow file is causing a null pointer deference or something can causing the reset vector on the Microcontroller to run.
… How big is your model? We’ve debugged the code quite a bit and the only issue we cannot directly prevent from happening is a stack overflow. If your model requires a huge stack beyond 10kb or so then this can happen.
Hi thanks for your quick reply,
Yeah its bigger than 10kb but I dont think that’s the case,because the only thing I’ve done was loading the model! Do you think overflow can happen with just loading a model? I’m new to tensorflow
Hi, the model size it’s what matters. Its what stack usage it causes…
Like, for example, Mobilenet is 1.5MB and it runs fine onboard.
Can you provide the code and the model?
Well i have some different models with a bit different structures i wanted to test them on the openmv cam so then i could decide which one i use!so i cant give you an exact number yet but their size varies from 100kb to 600kb! its a pretty wide range!but the error happens with all of them! and there is no specific code as i said i just try to load them !
Can you provide the model?