What is DUO?

The DUO is a platform consisting of Applications, Devices and Development Tools that enable people to create their vision. With the DUO M/MLX sensors we deliver compact stereo imaging.

Why is DUO better?

All DUO devices are precisely built and was designed to act as a professional grade 3D imaging solution. Here are some of our key differentiators:

– Compact Design
– Global Shutter
– Integrated MPU (6 DoF)
– Robust/Low Level API
– Broad OS Support
– Optimized Acquisition
– Versatile API/SDK and Docs

The DUO is proudly designed and built in the United States. We strive to provide researchers with stable and robust vision tools, enabling them to discover the future.

What is in the box?

Depending on which product you will receive the following in your package(s):

DUO DDK – DUO Development Kit

  • 1x DUO MLX Sensor
  • 1x DUO VPC Processing Unit
  • 1x DUO OS (microSD Card)
  • 1x VPC Power Supply
  • 1x Wifi Adapter
  • 1x HDMI Cable
  • 1x USB Cable (for MLX)
  • 1x Getting Started Card

DUO M – DUO mini Sensor

  • 1x DUO M Sensor
  • 1x USB Cable (Micro-B)
  • 1x Getting Started Card

DUO MLX – DUO mini lx Sensor

  • 1x DUO MLX Sensor
  • 1x USB Cable (Micro-B)
  • 1x Getting Started Card

If you are missing any of these components upon receiving/opening your package please contact order support at orders@duo3d.com.

How do I place an order?

To place an order, first you must create an account at the following link (https://duo3d.com/account). After you create an account and are logged in, you can go to the Products page and click “Order Now” to buy the product(s) you wish to order.

The items you select will be added to your cart and once you are ready to checkout simply click the checkout button on your Cart page. Upon checkout we will collect your shipping/billing details, calculate shipping costs and confirm the order items.

Once your order is confirmed you will be shown a “Thank you for your order” message and recieve the order receipt to your email address. If you require a custom invoice/receipt or have any troubles while ordering contact us directly at: orders@duo3d.com

What is the difference between the DUO M and MLX?

The key difference between the DUO M and the DUO MLX is that the MLX comes with a top board with the programmable LED array and integrated motion sensor (6 Degrees of Freedom). Both the M or MLX have the same footprint of 52x24mm with the MLX having a slighting taller profile of 13mm.

The DUO M is the core camera board with both image sensors and the USB controller while the MLX has the M base board as well as an additional top board to extend functionality. Beyond the top board the MLX also ships with IR bandpass filters that match the 850nm LEDs to assure only IR light is allowed into the sensors.

The M is best if you need to be able to experiment with different lens types or will only be using the sensor in a passive mode (no illumination).

The MLX is best in scenarios where you must control and limit the lighting conditions. This is common in most computer vision algorithms as it provides a simple way to guarantee consistent aspects of a scene.

To learn more about the differences make sure to check out the DUO Comparison page.

What languages do we support?

At the foundation of our development we use a pure C implementation for our Device API and related middleware (Dense3D). This allows for easy portability and compatibility when working with other languages or frameworks. Here is the list of supported languages/frameworks:


  • C/C++ (Official)
  • C# (Official)
  • Java via JNI (Community)


  • OpenCV (Official)
  • ROS (Community)
  • .NET (Official)
  • Qt (Official)

To learn more about getting started using your language or framework read the latest Integration Documentation article.

What ARM devices are supported?

We have tested the DUO on the following ARM devices:

– NVIDIA Jetson TK1
– Radxa Rock Developer Board
– DUO VPC Developer Kits

In general we recommend a minimum of 512mb of RAM with a processor over 1Ghz. Most ARM A9/A15 chips based devices be able to run DUO Devices in optimal form. We have performed many tests on different devices and for the sake of stability and usability we do not recommend using any lower end ARM developer boards such as the Raspberry Pi or Beaglebone. This may change in the future as we continue to optimize our Linux kernel and driver modules.

Will the DUO work with my code?

Yes. Since our API is exposed in from of C bindings, it is almost guaranteed to work in your workflow and development environment. Included with the SDK are simple examples that demonstrate DUO usage. On our API Docs page you can review the available methods and view code samples. If you are integrating the DUO into your project using OpenCV, QT or other frameworks reference our Integration Guides.

What is accuracy of the depth map?

In general we recommend using the DUO M/MLX sensors within 0.23-2.5m using the provided Dense3D and Calibration Applications. With this you should expect to receive around 0.1-0.3mm accuracy with epipolar lines that are aligned to within 0.05 pixels RMS error based on a stereo resolution of 320×240.

The resulting depth map accuracy or the sensor depends on several factors:

  • The algorithm used for extraction
  • The framerate, resolution and FOV
  • Illumination and texture
  • Distance from sensor
  • Calibration quality


Starting with the algorithm, we offer our Dense3D Middleware which uses a standard block matching method to perform stereo correlation. This is the most standard method used in the industry and acts as a base to test device functionality. For advanced users and developers please reference research into different stereo algorithms methods which can be used with DUO.

Resolution, Framerate and FOV

The increase of resolution will result in better accuracy at cost of performance. Framerate will not influence the resulting depth map accuracy, however, it will affect the illumination properties of a scene; the higher the framerate the more light required. By default on the DUO M/MLX devices we use a very high FOV (>150 degrees) if you are looking to achieve more accuracy at a distance consider using a more narrow field-of-view lenses as it improves the accuracy at long ranges.

Illumination and textures

For the stereo method to work properly it requires the objects in the scene are somewhat illuminated and have enough texture to differentiate themselves from other shapes in the scene. The DUO MLX ships with a programmable LED array which helps provide IR illumination even in dark scenes.

Distance from sensor

As we utilize the stereo effect our accuracy is directly tied to the distance from the imagers to the object. Thus we provide an accuracy calculator which provide both theoretical and applied accuracy measurements shown as a curve. The DUO M/MLX were specifically designed for short range sensing with the optimal accuracy with objects that fall within 0.25-1m.

Calibration quality

The proper calibration is critical in a stereo system. Please follow our Calibration Guide in full to get best results while calibrating your devices.