Step 1:   Use either Veloview to log data or the appropriate ROS driver available under the "Direct Driver Installation on Windows, Mac, or Linux" section below [Note: if you already have a Velodyne (.pcap) or ROSBAG (.bag) file, you may skip ahead to Step 3].

Step 2:  Generate a Velodyne (.pcap) or ROSBAG (.bag) file from a Velodyne Laser Scanner using the driver from Step 1 and angled as shown below under Suggestions for Data Collection.

Step 3:  Go to the Data Upload page, fill out the form, attach the laser scan file, and click on "UPLOAD”.  The Kaarta Upload Page will change to a Upload Success Page confirming your upload was received.

Step 4:  Kaarta will process your data, create a 3D model, and email you a link to download the point cloud usually within one business day. The processing is automated, but is reviewed prior to emailing you the link.

Direct Driver Installation on Windows, Mac, or Linux

The ROS Velodyne driver is an open source driver compatible with any system that has the Robot Operating System installed. The driver publishes messages to the /velodynepackets and /velodynepoints topics. Please log a rosbag (.bag) file containing /velodyne_packets (best) or /velodyne_points (suffiicient) that our 3D Model Generator can process.

How to Upload Laser Sensor Files TO KAARTA CLOUD

Supported Sensors and Data Formats
We currently support Velodyne’s HDL-64E, HDL-32E and Velodyne VLP-16 laser sensors. We will expand our service to support other 3D LiDAR brands soon.

We process Velodyne’s .pcap and ROSBAG .bag file formats. The .pcap files can be logged by the VeloView software provided by Kitware Inc. The ROSBAG .bag files can be logged by the ROS open source driver and contains packets (best) or points (sufficient) published to the /velodyne_packets topic (in the velodyne_msg/VelodynePacket data structure) or /velodyne_points topic (in the sensor_msgs/PointCloud2 data structure)  for the 3D model generator to process. In addition the following topics can also be processed by our 3D model generator:

- IMU with orientation information, topic name: /imu/data, topic type sensor_msgs/Imu

- IMU without orientation information, topic name: /imu/data_raw, topic type: sensor_msgs/Imu

- GPS data, topic name: /gps/data, topic type: sensor_msgs/NavSatFix


If you have files in the proper format, please go ahead and upload them here, otherwise, please follow the instructions to the right under "Direct Driver Installation on Windows, Mac, or Linux."


Velodyne’s .pcap format is preferred since the file sizes are smaller and the upload time is shorter.

What to Expect
Regardless of data file size, we will register 3D models containing one minute of LiDAR data for free. Please contact us at info@kaarta.com if you are interested in registering larger datasets.

What We Do with the Data
We will not share uploaded data with any third-party. However, we may contact you for permission to use interesting images for promotional purposes.

For best results please use our drivers as we have configured the ROS Velodyne driver to work with the supported sensor models. For the VLP-16, use this ROS driver and follow instructions inside the .zip file to install and operate the driver. For the HDL-32E, it is best to log the data using Veloview in order to capture the IMU data which we can use to improve your results. Veloview works well for the HDL-64E as well, although there is no IMU data. If you chose to use the ROS driver for the 32 or 64, please be sure it publishes the right topics mentioned above. For more information on topics please see the section on Supported Sensors and Data Formats.


Suggestions for Data Collection

Although not required, the following suggestions will improve the quality of the collected data and the final registered 3D point cloud. Other applications would share the same principles:

Handheld Based Data Collection

  1. Avoid turning and tilting the sensor at the same time. If you want to see higher on a surface, tilt the sensor while standing still. Try not to exceed 60 degrees.

  2. Rotate slower than 30 degrees/second when turning. If the sensor is tilted you will need to rotate more slowly.

  3. Smooth motions are better than jerky motions

  4. Backpack mounted sensors should be centered between the shoulders instead of off to one side.

  5. Scanning works best outdoors or large open spaces. It does not work well indoors in smaller spaces like houses, apartment buildings or offices.

Vehicle Based Data Collection

  1. Mount the LiDAR horizontally for recording the area around the vehicle or pitched down at a fixed angle between 10 and 15 degrees to record more of the area in front of the vehicle.

  2. The LiDAR should be mounted high enough to avoid sensor occlusion by the vehicle body.

  3. For better point cloud quality/density, vehicle motion should be slow and smooth (we suggest collecting data at less than 20 miles/hour). We also recommend starting the data collection for a few moments before initiating vehicle movement.

  4. Traversing the environment once slowly is better than traversing the environment multiple times quickly.

Aerial Data Collection

  1. IMPORTANT: Mount the LiDAR horizontally. Recommended flying heights vary up to 50 feet for the VLP-16 or 75 feet for the HDL-32E for generating decent point clouds. NOTE: Orienting the sensor on a UAV with the scanning axis vertically or tilted such that it cannot capture 3D data around the drone does not work with our registration software.

  2. The scan registration software also requires 3D structure in the environment and tends to fail when flying over flat fields or other unstructured environment.

  3. For better point cloud quality/density, aerial vehicle motion should be slow and smooth (we suggest collecting data at less than 20 miles/hour). The lower the flight, the higher the point density, but be careful of tree branches, or other items closer to the ground. We also recommend starting the data collection for a 20 seconds before initiating vehicle movement.

  4. Traversing the environment once slowly is better than traversing the environment multiple times quickly.

Data Visualization

After Kaarta Cloud has generated your 3D Model, Kaarta will send you a link to access your point cloud (.ply) files. Depending on the number of users, your file should will available within 1 to 2 business days.

You can use CloudCompare to open the generated point cloud (.ply) files. If you need a .las file for your processing pipeline, CloudCompare can open the .ply file and save it as a .las file. CloudCompare is open source software available for Windows, Mac, and Linux systems and is also available from source files.  

After loading the point cloud file, select the point cloud in the "DB Tree” toolbox (on the left hand side).  Then, select "Plugins > P.C.V. (Ambient Occlusion)" from the menu bar.  In the popup window enter "512" under "Count" and "512" under "Render context resolution" to create lighting/shadow effects.

Alternatively, you can select "Edit > Scalar Fields > Export Coordinate(s) to SF(s)" from the menu bar to colorize the point cloud based on Z-coordinates (elevation).  Next, find the "Properties"toolbox (below the DB Tree toolbox) and scroll down to the "Color Scale" section and change the "Current Color" option.  We recommend "Blue > White > Red".  You can also alter the color distribution by clicking and dragging the red or blue handles from the "SF display params" menu in the "Properties" toolbox.

To quickly colorize a point cloud you could also select (Eye Dome Lighting) EDL on the top menu bar.