Working with NumPy¶
Data structure of Open3D is natively compatible with NumPy buffer. The following tutorial generates a variant of sync function using NumPy and visualizes the function using Open3D.
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 | # examples/Python/Basic/working_with_numpy.py
import copy
import numpy as np
import open3d as o3d
if __name__ == "__main__":
# generate some neat n times 3 matrix using a variant of sync function
x = np.linspace(-3, 3, 401)
mesh_x, mesh_y = np.meshgrid(x, x)
z = np.sinc((np.power(mesh_x, 2) + np.power(mesh_y, 2)))
z_norm = (z - z.min()) / (z.max() - z.min())
xyz = np.zeros((np.size(mesh_x), 3))
xyz[:, 0] = np.reshape(mesh_x, -1)
xyz[:, 1] = np.reshape(mesh_y, -1)
xyz[:, 2] = np.reshape(z_norm, -1)
print('xyz')
print(xyz)
# Pass xyz to Open3D.o3d.geometry.PointCloud and visualize
pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(xyz)
o3d.io.write_point_cloud("../../TestData/sync.ply", pcd)
# Load saved point cloud and visualize it
pcd_load = o3d.io.read_point_cloud("../../TestData/sync.ply")
o3d.visualization.draw_geometries([pcd_load])
# convert Open3D.o3d.geometry.PointCloud to numpy array
xyz_load = np.asarray(pcd_load.points)
print('xyz_load')
print(xyz_load)
# save z_norm as an image (change [0,1] range to [0,255] range with uint8 type)
img = o3d.geometry.Image((z_norm * 255).astype(np.uint8))
o3d.io.write_image("../../TestData/sync.png", img)
o3d.visualization.draw_geometries([img])
|
The first part of the script generates a \(n \times 3\) matrix xyz
.
Each column has \(x, y, z\) value of a function \(z = \frac{sin (x^2+y^2)}{(x^2+y^2)}\).
\(z_{norm}\) is normalized map of \(z\) for [0,1] range.
From NumPy to open3d.PointCloud¶
25 26 27 28 | # Pass xyz to Open3D.o3d.geometry.PointCloud and visualize
pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(xyz)
o3d.io.write_point_cloud("../../TestData/sync.ply", pcd)
|
Open3D provides conversion from NumPy matrix to a vector of 3D vectors. By using Vector3dVector
, NumPy matrix can be directly assigned for open3d.PointCloud.points
.
In this manner, any similar data structure such as open3d.PointCloud.colors
or open3d.PointCloud.normals
can be assigned or modified using NumPy. The script saves the point cloud as a ply file for the next step.
From open3d.PointCloud to NumPy¶
30 31 32 33 34 35 36 37 | # Load saved point cloud and visualize it
pcd_load = o3d.io.read_point_cloud("../../TestData/sync.ply")
o3d.visualization.draw_geometries([pcd_load])
# convert Open3D.o3d.geometry.PointCloud to numpy array
xyz_load = np.asarray(pcd_load.points)
print('xyz_load')
print(xyz_load)
|
As shown in this example, Vector3dVector
is converted into a NumPy array using np.asarray
.
The tutorial script prints two identical matrices
xyz
[[-3.00000000e+00 -3.00000000e+00 -3.89817183e-17]
[-2.98500000e+00 -3.00000000e+00 -4.94631078e-03]
[-2.97000000e+00 -3.00000000e+00 -9.52804798e-03]
...
[ 2.97000000e+00 3.00000000e+00 -9.52804798e-03]
[ 2.98500000e+00 3.00000000e+00 -4.94631078e-03]
[ 3.00000000e+00 3.00000000e+00 -3.89817183e-17]]
Writing PLY: [========================================] 100%
Reading PLY: [========================================] 100%
xyz_load
[[-3.00000000e+00 -3.00000000e+00 -3.89817183e-17]
[-2.98500000e+00 -3.00000000e+00 -4.94631078e-03]
[-2.97000000e+00 -3.00000000e+00 -9.52804798e-03]
...
[ 2.97000000e+00 3.00000000e+00 -9.52804798e-03]
[ 2.98500000e+00 3.00000000e+00 -4.94631078e-03]
[ 3.00000000e+00 3.00000000e+00 -3.89817183e-17]]
and visualizes the function:
From NumPy to open3d.Image¶
2D Numpy matrix can be converted image. The following example converts z_norm
into open3d.Image,
visualize the image using draw_geometries
, and save the image as a png format file.
39 40 41 42 | # save z_norm as an image (change [0,1] range to [0,255] range with uint8 type)
img = o3d.geometry.Image((z_norm * 255).astype(np.uint8))
o3d.io.write_image("../../TestData/sync.png", img)
o3d.visualization.draw_geometries([img])
|
Note
The conversion supports uint8
, uint16
, or float32
with c_type storage (default NumPy behavior),
dim=2`` (width * height) or dim=3 (width * height * channel).