Train a neural network

Hello everybody, I’m Patrick and I just started using an Open MV3 R2.

First of all, the purpose of my work is to detect a particular object and register its movements (quantify them and record them, writing the output to a file).

I am quite familiar with Python, but I never trained a neural network before.
I saw that there is a tutorial on Github, but apparently is only for Ubuntu users.

I have got some questions.

  1. Do you think it’s possible create a NN on Windows?
  2. Should I create the .network file before on a Jupiter notebook and then load it into the OpenMV environment ?
  3. How does the path works (for external files) ? For instance if I want to save an image or to read it, where should I locate it ?


    Thank you very much for any indication that you’ll provide to me.

Hi Patrick

  1. Do you think it’s possible create a NN on Windows?

Yes, if you can setup Caffe on Windows you should be able to produce the same results. However I would just use Ubuntu (things just work on Linux)

  1. Should I create the .network file before on a Jupiter notebook and then load it into the OpenMV environment ?
  2. How does the path works (for external files) ? For instance if I want to save an image or to read it, where should I locate it ?

Please see the README file here, it covers the whole process in details (This will be merged into the project soon):

https://github.com/openmv/openmv/tree/master/ml/cmsisnn