Visualizations in 3D

The Ouster Python SDK provides two visualization utilities for user convenience. These are introduced briefly below.

Visualization with Ouster’s simple-viz

Ouster’s OpenGL-based visualizer allows for easy visualization from pcaps and sensors on all platforms the Ouster SDK supports.

The default Ouster simple-viz visualizer view includes two 2D range images atop which can be cycled through the available fields, and a 3D point cloud on the bottom. For dual returns sensors, both returns are displayed by default.

Ouster simple-viz visualization of OS1 128 Rev 6 sample data

The visualizer can be controlled with mouse and keyboard:

Keyboard controls:

Key	What it does
o	Toggle on-screen display
p/P	Increase/decrease point size
m	Cycle point cloud coloring mode
b	Cycle top 2D image
n	Cycle bottom 2D image
R	Reset camera
e/E	Increase/decrease size of displayed 2D images
'/"	Increase/decrease spacing in range markers
w	Camera pitch up
s	Camera pitch down
a	Camera yaw left
d	Camera yaw right
1	Toggle first return point cloud visibility
2	Toggle second return point cloud visibility
0	Toggle orthographic camera
=/-	Dolly in/out
(space)	Toggle pause
./,	Step one frame forward/back
ctrl + ./,	Step 10 frames forward/back
>/<	Increase/decrease playback rate (during replay)
shift	Camera Translation with mouse drag
shift-z	Save a screenshot of the current view
shift-x	Toggle a continuous saving of screenshots

To run the visualizer with a sensor:

$ simple-viz --sensor $SENSOR_HOSTNAME

This will auto-configure the udp destination of the sensor while leaving the lidar port as previously set on the sensor.

To run the visualizer with a pcap:

$ simple-viz --pcap $SAMPLE_DATA_PCAP_PATH --meta $SAMPLE_DATA_JSON_PATH

Visualization with Ouster’s viz.PointViz

Please refer to PointViz Tutorial & API Usage for details on extending and customizing viz.PointViz.

Visualization with Open3d

The Open3d library contains Python bindings for a variety of tools for working with point cloud data. Loading data into Open3d is just a matter of reshaping the numpy representation of a point cloud, as demonstrated in the examples.pcap.pcap_3d_one_scan() example:

# compute point cloud using client.SensorInfo and client.LidarScan
xyz = client.XYZLut(metadata)(scan)

# create point cloud and coordinate axes geometries
cloud = o3d.geometry.PointCloud(
    o3d.utility.Vector3dVector(xyz.reshape((-1, 3))))  # type: ignore
axes = o3d.geometry.TriangleMesh.create_coordinate_frame(
    1.0)  # type: ignore

The examples.open3d module contains a more fully-featured visualizer built using the Open3d library, which can be used to replay pcap files or visualize a running sensor. The bulk of the visualizer is implemented in the examples.open3d.viewer_3d() function.

Note

You’ll have to install the Open3d package from PyPI to run this example. Note that as of version 0.13.0, binaries are not yet provided for Python 3.9 or ARM systems.

As an example, you can view frame 84 from the sample data by running the following command:

Linux/macOSWindows x64

$ python3 -m ouster.sdk.examples.open3d \
    --pcap $SAMPLE_DATA_PCAP_PATH --meta $SAMPLE_DATA_JSON_PATH --start 84 --pause

You may also want to try the --sensor option to display the output of a running sensor. Use the -h flag to see a full list of command line options and flags.

Running the example above should open a window displaying a scene from a city intersection, reproduced below:

Open3D visualization of OS1 sample data (frame 84). Points colored by SIGNAL field.

You should be able to click and drag the mouse to look around. You can zoom in and out using the mouse wheel, and hold control or shift while dragging to pan and roll, respectively.

Hitting the spacebar will start playing back the rest of the pcap in real time. Note that reasonable performance for realtime playback requires relatively fast hardware, since Open3d runs all rendering and processing in a single thread.

All of the visualizer controls are listed in the table below:

Open3d Visualizer Controls

Key	What it does
Mouse wheel	Zoom in and out
Left click + drag	Tilt and rotate the camera
Ctrl + left click + drag	Pan the camera laterally
Shift + left click + drag	Roll the camera
“+” / “-“	Increase or decrease point sizes
Spacebar	Pause or resume playback
“M”	Cycle through channel fields used for visualization
Right arrow key	When reading a pcap, jump 10 frames forward

Visualization with Matplotlib

You should have defined source using either a pcap file or UDP data streaming directly from a sensor, please refer to Developer’s Quick Start with the Ouster Python SDK for introduction.

Note

Below pictures were rendered using OS2 128 Rev 05 Bridge sample data.

Let’s read from source until we get to the 50th frame of data:

from contextlib import closing
from more_itertools import nth
with closing(client.Scans(source)) as scans:
    scan = nth(scans, 50)

Note

If you’re using a sensor and it takes a few seconds, don’t be alarmed! It has to get to the 50th frame of data, which would be 5.0 seconds for a sensor running in 1024x10 mode.

We can extract the range measurements from the frame of data stored in the LidarScan datatype and plot a range image where each column corresponds to a single azimuth angle:

range_field = scan.field(client.ChanField.RANGE)
range_img = client.destagger(info, range_field)

We can plot the results using standard Python tools that work with numpy datatypes. Here, we extract a column segment of the range data and display the result:

import matplotlib.pyplot as plt
plt.imshow(range_img[:, 640:1024], resample=False)
plt.axis('off')
plt.show()

Note

If running plt.show gives you an error about your Matplotlib backend, you will need a GUI backend such as TkAgg or Qt5Agg in order to visualize your data with matplotlib.

Range image of OS2 bridge data. Data taken at Brooklyn Bridge, NYC.

In addition to viewing the data in 2D, we can also plot the results in 3D by projecting the range measurements into Cartesian coordinates. To do this, we first create a lookup table, then use it to produce X, Y, Z coordinates from our scan data with shape (H x W x 3):

xyzlut = client.XYZLut(info)
xyz = xyzlut(scan)

Now we rearrange the resulting numpy array into a shape that’s suitable for plotting:

import numpy as np
[x, y, z] = [c.flatten() for c in np.dsplit(xyz, 3)]
ax = plt.axes(projection='3d')
r = 10
ax.set_xlim3d([-r, r])
ax.set_ylim3d([-r, r])
ax.set_zlim3d([-r/2, r/2])
plt.axis('off')
z_col = np.minimum(np.absolute(z), 5)
ax.scatter(x, y, z, c=z_col, s=0.2)
plt.show()

Point cloud from OS2 bridge data with colormap on z. Data taken at Brooklyn Bridge, NYC.

You should be able to rotate the resulting scene to view it from different angles.

To learn more about manipulating lidar data, see:

• Staggered vs Destaggered 2D Representations

• Projecting into Cartesian Coordinates

• Working with 2D and 3D Representations Simultaneously