Sensor configurations

Depthkit Studio supports up to 10 Azure Kinects. Sensor configurations are customizable based on your creative needs, capture space, and subject matter. Below, explore specifications for a 3, 5, and 10-sensor setups as well as best practices for assembling and positioning your sensors on stage, checking the sensors frames, etc.

📘

In any sensor configuration, the closer the sensor is to the subject the higher the quality.

In this page

3-Sensor configuration →
5-Sensor configuration →
10-Sensor configuration →
Positioning Sensors →


3 Sensors

Sensor coverage: approximately 270º of your subject.
Calibration time: approximately 20 minutes

The three-sensor setup is a configuration that's quick and easy to set up, and is ideal for publishing to mobile.

The configuration consists of three sensors in a front-focused orientation. As seen below, one central sensor is placed at the height of the intended final viewer that is oriented directly at the front of the talent. Two sensors are positioned on either side of that central sensor. These two side sensors can be positioned slightly lower to see under sides of chin. This will allow for focus on the face as the prime area of focus. For best results, all three sensor perspectives should have the full body (or intended subject) within the frame.

3 Sensor capture results

5 Sensors

Sensor coverage: 360º of your subject
Calibration time: approximately 30-60 minutes

A full body capture configuration designed for a lightweight setup that captures the detail of your subject.

This design consists of five sensors. Four are arranged in a square formation. The fifth sensor is located at the front, positioned closer to your subject's face in order to capture higher quality facial detail.

5 Sensor capture results.

10 Sensors

Sensor coverage: 360º of your subject
Calibration time: approximately 1 hour+

While this configuration is flexible, we find that the one pictured above is an ideal distribution of sensors that solves for occlusion and minimizing the angles between sensors overall.While this configuration is flexible, we find that the one pictured above is an ideal distribution of sensors that solves for occlusion and minimizing the angles between sensors overall.

While this configuration is flexible, we find that the one pictured above is an ideal distribution of sensors that solves for occlusion and minimizing the angles between sensors overall.

This sensor configuration represents the highest possible number of sensors that our system can currently support and creates high-quality results with a very low incidence of occlusions.

It is also configurable, enabling a studio to create capture volumes ranging from 3ft-6ft or more while still maintaining quality comparable to the results below. However, as a general rule the closer the sensor is to the subject the higher the quality.

The primary strength of this configuration is the excellent coverage and minimal occlusions which can support complex performances, gestures or productions that require multiple subjects. That coverage is most helpful in areas where the sides of the body are hidden by the arms, or the interior surfaces of the legs may be hidden by the opposing leg. The resulting reconstruction will also be more detailed and accurate overall.

📘

Automatic Sample Filtering

As depicted in the above video, previous version of Depthkit required manually covering oblique sensors to avoid capturing samples with high amounts of error.

In the current version of Depthkit, this is unnecessary, as samples captured from oblique angles are automatically filtered out.

Our recommended sensor configuration puts the sensors quite close to the subject to maximize the quality produced from the sensors – closer is always better with depth sensors – but you can increase the diameter of your rings and distance to subject as you need in order to create a larger central capture volume.

Alternatively, you can also set your sensors to a wide or narrow field of view in the Depthkit software interface after you have calibrated which will allow you to dilate your capture region. These capture volumes can be very clearly understood when installing your sensors – simply ensure that your subject remains inside the frames of all of your surrounding depth and color streams as seen in the interface.

Positioning Sensors

The closer you place your sensors to the subject, the more detailed your capture will be, but this may also reduce the size of the usable volume you are able to capture.

Azure Kinect depth camera vignette

Keeping all of the captured movement inside the unshaded area prevents artifacts generated by areas of the color frame without valid depth data. Anything captured in the semi-shaded portion of the frame will have valid depth data, but may produce other artifacts near the corners of the frame.

Sensor orientation

You can shoot with your sensor positioned horizontally (A) or vertically (B). Shooting vertically can be best suited to a standing, full body posture due to the shape of the Azure Kinect's depth camera vignette. Depending on the motion, it may be beneficial rotate the sensor, as seen in the Depthkit Edit workspace on the right. In either case, rehearse your blocking to make sure that your subject stays within both the color and depth frames.

Vertical orientation, as seen in Depthkit.Vertical orientation, as seen in Depthkit.

Vertical orientation, as seen in Depthkit.