Frequently Asked Questions
How can I generate point clouds from the recorded data?
To ensure real-time performance during data recording, point clouds are not saved directly. However, you can generate point clouds from the recorded data afterward. Follow this example to create point clouds offline:
How can I generate ROS2 bags from the recorded data?
ROS2 bags are not directly recorded to ensure optimal performance during data capture. However, you can generate ROS2 bags from the recorded data afterward. Follow this example to generate ROS2 bags from recorded data:
The 3D bounding boxes in nodar_viewer are not aligned with the ground
The 3D bounding boxes in nodar_viewer are often not aligned with the ground
and/or have an offset with respect to the ground when the camera axis is not parallel with the ground plane.
Most issues are fixed by adjusting bar_pitch in the master_config.ini.
In nodar_viewer, you can adjust this by going to
Viewers Tab -> Point Cloud -> ⚙️ -> Bounding Boxes -> Display Boxes [✅]
Then tune the Box Offset parameter to adjust the relative position of the 3D bounding boxes along the y axis.
The distance to an object gives me the wrong value
During shipment and large shock events, the absolute range calibration can sometimes be affected.
To recalibrate the range measurements, run nodar_viewer, and go to
Viewers Tab -> Initial Calibration -> ⚙️ -> Range Calibration -> Add Region
Then select a target in the image with a known distance, enter the distance to the target, and click Submit.
When the calibration is complete, recheck the distance to the target in the Colorized Depth Map
(click a point in the image so see the distance).
If the ranges still seem to be wrong, you should try to select a different target and
repeat the range calibration routine.
An ideal target would be relatively large and far away, that is, more than 30 m or 100 ft away.
The detection is cut off in the bottom left corner of the Colorized Depth Map
Shadows around objects and missing returns in the bottom left corner are expected. The bottom left corner having no returns is often due to the field of view. Hammerhead can only return depth estimates in regions where the fields-of-view of the left and right cameras overlap. The bottom left corner of the left image is often cut off because that portion of the image isn't visible in the right image.