I’m trying to train a Detectnet_V2 model using Isaac sim to generate data. I have 16 classes but I don’t know how many images I choose for each one I’ve tried 4000 per each but it gave an over fitted model for two classes an a useless one for 16 classes, should I consider making small objects larger even if I don’t respect the proportions?
and how about spawning many objects at the same time?
Hi, were you able to make any progress with finding out how much data is enough? I’ve gotten good results on the test dataset but the object detection doesn’t work that well in my Unity3D simulations.