Jetson Nano AI

[ali.yusuf]

Hi,
I am new to the whole field of programming and AI.
I want to teach my Jetson Nano to identify traffic signs. I have found a suitable source code for JupyterLab Notebooks. However, I execute the code every time the Jetson is switched off and on. This means that the learning process is done every time. And each time new random data from the existing database is used for teaching. So the result is always different (for the same object).
Is there a possibility that I can somehow save what I have learned and when I switch on the JetsonNano again, I can continue with what I have learned once?
If so, how can I realize this ?

Many thanks and best regards

[dusty_nv]

Hi @ali.yusuf, if you are training a neural network model, typically you would save a ‘checkpoint’ of your model that could be loaded later. For example, this is how you would do it in PyTorch:

• https://pytorch.org/tutorials/recipes/recipes/saving_and_loading_a_general_checkpoint.html
• https://pytorch.org/tutorials/beginner/saving_loading_models.html

Other frameworks such as TensorFlow have a different mechanism for saving/loading model checkpoints that you can find in that framework’s documentation and tutorials.