please give me some advice

[LeoKingBUAA]

I'm a newcomer to deep learning, and now I have multiple semantic road vectorized point clouds that need non-rigid registration.Could you please provide guidance on how to adjust the model for improved matching results?