computation capacity of jetsonano

Hi I’m trying to carry out a platooning project with two Jetsonnano.
I wonder if Jetson Nano can do deep learning operations without a server computer and send data to follow cars while driving autonomously.

Hi e1205703, yes you can do some level of training onboard Jetson Nano. Larger networks and datasets are better suited to training on a workstation with discrete GPU(s) or the cloud, however you can do some transfer learning onboard Nano (in particular, we have found PyTorch to be efficient enough on memory usage).

These following deep learning examples perform the training onboard Nano using PyTorch:

Thank you for your answer. Dusty
May I ask you how to the line dectection on CNN at Jetson Nano?
And sent github re of the training we can learn the line?

If you mean line following, I would check the JetBot or JetRacer projects, which can be trained to follow paths.

The DLI Nano course includes training a regression network to predict the locations of facial features. In theory you could modify the code to detect the line instead.