OpenMV H7 Plus Usage of 32 MB SDRAM

Hi! Im trying to load a model which is 7.4MB. I got an error that “Memory Allocation” expectedly. Then I ve reasearched this error and I saw an advice that "Use ‘load_to_fb=True’ ". I ve tried it but whenever I try to load the model, the OpenMV is restarting itself without any exceptions, errors etc.

This is the line of code that program stucks and OpenMV restarts itself:

net = tf.load("model.tflite", load_to_fb=True)

How can I load my 7.4MB model to 32 MB SDRAM?

It’s because the model has some tensorflow ops that we don’t support. Check that your model against these ops: tensorflow-lib/libtf.cc at master · openmv/tensorflow-lib · GitHub

I realize this behavior isn’t great. However, the ACTUAL tensorflow library from GOOGLE segfaults when it has an error…

Thanks a lot! Also, do you know any template that i can train then use on H7 Plus without Edge Impulse?

There’s an example here:

1 Like

Thank you @iabdalkader but this code is classification example. Could you please advice me any object detection training sources?

That’s the only example that I know of.