The example train_RT3D.py scripts for Figure_3 and Figure_4 are almost identical. Should refactor into general training function