That's it! OK, maybe I should have said step instead of steps. It may not be that simple in all cases, but it's pretty simple generally speaking. While you can use "Character-based" sizes like you did ...

In brief, I have some problems transforming predictions on suggested frames into labels. When I tried to do so, the GUI would stuck. I am using 1.6.0a3 and I used trained top-down models to infer on ...