prepare dataset from h5 files with classification label per point

[h612]

Hi, I'm new to pointcloud and need some help
I'm trying to use PointNet on my dataset which is of following nature:

• I've h5 files (each row has rgb,xyz data)
• Also, I've a .bin file which has each points' Ground Truth.
• The dataset is pointcloud of pipelines and seabed. Thus each point has its own label in 1 file.

dataset is as:
[[x1,y1,z1],
[x2,y2,z2],
[x3,y3,z3],
[x4,y4,z4]..]
`label is as`
[1,
0,
0,
1,
..]

I've seen couple of examples, but each is labelled per entire pointcloud, e.g. one folder CHAIR has Pointcloud of chairs only.
I'm stuck at this point of data preparation in this case.
I was trying to do a test train split to get test and train data for Pointnet, but I'm clueless.
Please help.
Thanks