I have a custom human detection model that I trained using Yolo11 in Ultralitics HUB. Using pytorch in TD is relatively easy and it worked for me. However, I would like to use the native TD BodyTrack CHOP node. I exported the model to the Nvidia Tensor RT format. As a result I got a file with .engine extension. I see that Bodytrack CHOP uses the models located in the /Config/Models directory to detect people. However, these models have the extension engine.trtpkg. As a result, my question is very simple - is there any way to use my models with bodytrack chop? Thanks.
Interesting idea. I can’t really say for sure, but I doubt you could replace the model.
The Body Track CHOP is just a wrapper for the Nvidia Maxine AR library, so we don’t actually do any of the lower level interactions with the model ourselves and so we don’t have any control over what model is loaded.
If you want, you could try posting to the Nvidia github or developer forum to see if they have any more information.