Supported by

We will be at FUTURECOM 2017 inside Qualcomm space. Come visit us!

What is it for?

DragonWally is an embedded solution for image processing based on the Snapdragon 410E platform.

... ...

Object detection

The detection and recognition algorithms can be trained for identification of any type of object, besides estimating their distances to the camera.

People counting

With a 3D processed image, the task of people counting becomes easy and more accurate. DragonWally can be used for people counting, analyzing the flow of people and even customer satisfaction assessment (using sentimental face detection).

Access control

Everyone has unique face features and, with accurate face detection algorithms, DragonWally can be used for access control and tracking of people in buildings.

Driver identification and safety

DragonWally can be used inside vehicles, for driver and passenger identification, supervision of the use of seat belts, detection of driver's drowsiness, amongst many other safety oriented applications.


Two high-resolution cameras take pictures of the same scene from different points of view. An epipolar geometry algorithm then calculates the distance of the detected objects, based on subtle disparities between the two images.

DragonBoard 410c SBC

The DragonBoard 410c SBC from Arrow/Qualcomm provides the computing power and multimedia capabilities to handle all features of computer vision with stereo cameras.

Mezzanine DragonWally DW0

We've design a custom mezzanine (DragonWally DW0) that allows efficient interfacing of two high-resolution cameras.

OpenCV and Python

Python and OpenCV streamline the tasks of image capturing, image pre-processing and face detection.

AWS Rekognition

AWS Rekognition API from Amazon Cloud Solutions provides tbe machine learning algorithms used by DragonWally, resulting in high accuracy of face recognition.

Camera Mezzanine

DragonWally DWO Mezzanine add-on board for DragonBoard 410c SBC


The DragonWally DW0 Mezzanine with two OV5645 (OmniVision) camera modules enables rapid prototyping and development of computer vision applications using the DragonBoard410c SBC. The direct interfacing of cameras through MIPI-CSI signals allows for a robust implementation of embedded computer vision processing, eliminating the need of external cameras (using USB ou Ethernet).

The mezzanine board has an additional USB port, equipped with overcurrent protection (500mA), ESD protection of the D+/D- signals and power On/Off control through GPIO port of the DragonBoard 410c.


The mezzanine was designed with the purpose of supporting computer vison applications in the DragonBoard 410c SBC.

  • Conforms to the 96Boards Consumer Edition especification
  • Supports DragonBoard 410c
  • Direct MIPI-CSI camera interfacing
  • Supports two 5 Megapixel cameras
  • Supports Linux (Linaro Debian) OS
  • Supports OpenCV e Python for rapid computer vision application development
  • 1 USB port

Product Specifications

Datasheet download: ENG / PT_BR

  • 1 DragonWally DW0 Mezzanine PCB
  • 2 OV5645 camera modules

Customization Services and Product Design

We offer customization services of this reference design for new applications meeting additional client requirements.

Our engineering team has solid experience with hardware and software design for the Snapdragon 410E platform.

Inquires and orders can be made by email:


How it works

Our Team


Cezar Menezes

Electronic Engineer

Solid background in designing M2M (IoT) electronic hardware and firmware for Power Utilities (Smart Grid) and Telematics (AVL, Fleet Management).
Graduation: BSEE degree from Escola Politécnica/USP (São Paulo, SP, Brazil) in 1992.


Cleber Picolo

Computer Scientist

Strong experience in Full Stack development, SCADA Applications and Machine Learning.
Graduation: CS degree from Unicamp (Campinas, SP, Brazil) in 2010.

Watch our interview to DragonTalks in Qualcomm's facebook

Watch interview

Leave a message

We have received your mail, We will get back to you soon!

Sorry, Message could not send! Please try again.