I noticed that you can setup your own custom models in the menu, but I'm curious how would one train such for Hotscreen? I imagine it's loads of work (using Ultralytics and YOLOv11 for example) but is it feasible for an ordinary casual user to do it? Is it all just too much to explain?

I would like to provide a python script that can finetune a model with the images of your own dataset. You will still need to know how to use python and the most work would be to labelize your dataset which takes time.

Also, some new transformers models could work better than the yolos, so I will try them (this month) before creating a tutorial.

That'd be much appreciated, will patiently wait for that then. Good job on the software and thanks for the active development and communication :)