Real Time Object Recognition on Jetson Nano

I have trained a CNN in colab with a dataset of 10500 images of sign language. Now I want to import this model into the Jetson Nano and be able to make recognition in real time (20-50 fps).

What I want for you is to construct a pipeline so I can make modifications to the CNN in colab, then import this model and test it in real time into the Jetson Nano.

I only have the CNN in colab, the 10500 images in jpg format (rgb and grayscale 176 x 144 px ) and a csv which represent those grayscale images in vectors.

I have the 4gb Jetson Nano and a raspberry camera with Jetpack already installed.

Could you help me with that?

Thanks in advance.