Train test splitting question

[HRHLALALA]

Hi, do you train on 4 datasets and test on one dataset? It seems you are using 70% of data to train and 30% data for testing from your code. Also, your position normalisation uses same minimum and maximum value on all scenes .

If not, can you describe how you normalise the data for mixed scenes? Thanks.