Best approach to detect 2 classes in images on a Nano 2GB?

Hi All

We know 2GB is not a lot. So how to use them best? I want to detect 2 classes: boxes and covers as shown in image below. It works great on my PC but we’re running out of memory on the Nano 2GB using Pytorch and YoloV5s.
What would you recommend?



You can check our jetson_inference sample.
The sample converts a detector into a TensorRT engine which requires less memory.