To kick the can down the road some more: Su is a 32 bit app now, but will probably chage when MS goes to 64 bit in total ~Windows 10. That ( 32 B) means SU trades precision for range of numbers and the present precision is 7 digitis +. Therefore given the inference engine must have some tolerance for its delecrations it reports it is not always correct so, it would help you to spend some time reading and watching videos on its ( inference engine) use . This model will be a good learning experience for you. There are some small dimensions in you model and some times scaling it up say by ~ x10=>x100 can help and then scale back down when you are done.
Gday
The tolerances are not OS related and would not change with 64bit.
I will read more about (inference engine)… for me, I have applied exact measurements using the reference lines… but anyway, I should revise it well using your comments.
Thanks.
Tolerance may not change, unless Su wants to undo some of the games Google played with conversion I responded too several years ago. Sixty four bit gives you less graduality on the number axis and greater precision vs the ~24 bits you get with a 32 bit float.
1 Like