How to access points from o3d.t.io.read_point_cloud?

[nimafo]

Dear community,
I am trying to access points in the form of Numpy XYZ array from the following block of code for the main points cloud(point_cloud) and the boundary points(boundaries) which have the same data structure. How can I do that?

import open3d as o3d

# Load or create a point cloud
ply_file_path = 'cluster_4.ply'
point_cloud = o3d.t.io.read_point_cloud(ply_file_path)
point_cloud.estimate_normals()
radius = 0.05
max_nn = 15
boundaries, mask = point_cloud.compute_boundary_points(radius, max_nn)

when running the variables I get the following:

PointCloud on CPU:0 [74919 points (Float32)].
Attributes: normals (dtype = Float32, shape = {74919, 3}).

[saurabheights]

If you are trying to convert from numpy to PointCloud, you can try:-

http://www.open3d.org/docs/release/python_api/open3d.utility.Vector3dVector.html

Note: you can add color of pointcloud in same way as points.

import open3d
import numpy as np

pcd = open3d.geometry.PointCloud()
np_points = np.random.rand(100, 3)
np_colors = np.random.rand(100, 3)  

# From numpy to Open3D
pcd.points = open3d.utility.Vector3dVector(np_points)
pcd.colors = open3d.utility.Vector3dVector(np_colors)

open3d.visualization.draw(pcd)

Result:-

[nimafo]

I am actually trying to convert o3d.t.io.read_point_cloud(ply_file_path) to numpy array.

[ssheorey]

HI @nimafo you can use the numpy() function on Tensors to convert from an Open3D tensor to a Numpy array.

# Access the `positions` Tensor property of the `point` data structure in a point cloud.
npy_pts = point_cloud.point.positions.numpy()   

http://www.open3d.org/docs/latest/python_api/open3d.core.Tensor.html?highlight=numpy#open3d.core.Tensor.numpy