Hello, if I want to run my own sequences, do I need to pretrain using my image datasets or I can just run tracking? Thank you!