Jetson nano Training

Hi all

I would just like to say hello and introduce myself, I work for a very large multinational company and I purchased a Jetson Nano developer kit.

I am unsure if I should post this here, so please correct me and explain where I should post if not correct.

I am very interested in the vision capability and the ability to recognise objects and detect.

I am a complete novice and have been working my way through the 2 days to a demo tutorials. I have managed to get it all up and running, I just need to start training my own model. I would like to start with something simple and be able to detect if a bottle is stood upright or has fallen over.

I want to be able to use this for machine infeed/outfeed areas so we can extract the data.

I am unsure how I would go about or code the jetson to write what it sees to an excel file? I can then export the excel file to Power Bi or something similar to write reports automatically.

Can anyone offer any advice on where I would start, or point me in the right direction of where I might already find this information so I can train myself.

Thanks

Mark

Hi Mark, welcome to the community! First I would recommend running through these training examples on your Nano:

https://github.com/dusty-nv/jetson-inference/blob/master/docs/pytorch-transfer-learning.md

Then, you can collect your own dataset using the camera-capture program that’s covered in the tutorial. Your dataset can have two object classes: bottle up, and bottle down. So you will want to capture many images of each scenario, then re-train the network on your bottle dataset. Good luck!

Hi

Thankyou for the speedy response… i am doing this as we speak… hopefully it will work…

How do I program the nano to export data to an excel file…

for instance every time the bottle is fallen, time, date etc…

Is there any where i can read for this?

Thanks

Mark

It would probably be easiest to export the data from Python to C++ to a CSV or text file as opposed to an Excel XLS (and Excel can then still import CSV).

What you would want to do, is modify the imagenet-camera program, so that after it runs the classification it dumps the result to the CSV or text file. If you are using Python, you can use the Python csv module to write the CSV file. And here is a simple example of writing CSV file from C++.

Thank you so much

Mark