I nearly added this feature in the last update, but I was afraid it would break things for some people. In the next update, I will add a setting to select which gpu (or cpu) is used to run the detection model.
I appreciate it!
