Apperently it didn't had to do with the way I tried to save and import the network but with the version I used to save it and than load it again. You see, I trained the network on a workstation which has multiple GPUs, more RAM and a more powerful CPU than my personal laptop at work. But as testing the network doesn't really require a powerful PC, I transfered the trained network files to my own laptop to test the network. Training the network was done in Matlab 2019a while I have 2018b on my laptop. This resulted in the network to be empy as I imported it matlab 2018b.
Moral of the story is to not use different versions of matlab in the same project!
But thanks to everyone for their sugestions.