Parameters in normalization

[yilutum]

GoalFlow/navsim/agents/goalflow/goalflow_model_traj.py

Lines 520 to 521 in 411fdbd

	x_scale = 60
	y_scale = 15

Hi GoalFlow team,

Thank you for this excellent work. I have a question about the trajectory normalization in the code: how were x_scale and y_scale determined?

If I want to train on a new dataset, what would you recommend I do to choose appropriate scaling values (e.g., compute from dataset statistics, follow a fixed physical range, or reuse the same values)?

Thank you in advance.