Hi,
I have one question: With distinct trained models with a “freezed” backbone, is it possible to use just one instance of the backbone and the resulting features on all the heads from all the models, like shown in the diagram?
Thanks for the help!