Volumetric Capture with Azure Kinect vs Kinect v2
Technical specs and why everyone working with volumetric video should get excited.
Probably the most anticipated products of 2019 for all computer vision and volumetric capture users is the new Azure Kinect camera from Microsoft corp. As it’s predecessors Kinect v2 for Windows and Kinect 360 will be used to obtain real-time depth data.
Why it is better than ever?
The new hardware comes with 12MP CMOS rolling shutter sensor. This alone gives a vast freedom for developers to build versatile applications including volumetric video capture. Kinect azure enables variety of frame rates and resolutions from both of internal RGB and depth sensors. It stands out with synthetic RGBD resolution of 4096×3072 that generates an astonishing 12 million points every 66.6 milliseconds. In comparison to Kinect v2 which tops only 2 million points per frame. Thats 6 times more points which makes it great for capturing good quality volumetric video. Check out the specs yourself here. Other super handy feature that is worth mentioning is the hardware synchronisation via 3.5 mm jacks allowing multiple Azure Kinect camera frames to be synced together.
Portability factor
The old Kinect v2 for Windows was designed as an XBOX camera for body tracking experiences and it was meant to be stationed in one place. It had a mass of 960 grams with dimensions of 249 x 66 x 67 mm. In addition to it’s bulky appearance it came with two cables which needed to be connected to a Windows PC. On the contrary the new Azure Kinect is at half in size and weight. It comes with a slick USB-C type cable. However, when using Azure Kinect for volumetric capture we soon realised that most of the time thats not enough and the cable needs to be extended. We recommend using NEWNEX cables. More information on hardware needed for volumetric video capture can be found here.
Software: is it enough for volumetric capture?
Microsoft states on their website that Kinect Azure DK hardware and Developers Kit are not compatible with Kinect v2. With a new Kinect camera new Kinect Azure Viewer application was introduced. It mainly enables to inspect the depth and RGB capabilities but that is almost it. Apart of that there are virtually no solution to plug-and-play Kinect cameras. Regarding Kinect v2 there are open source code to play around built by the community of developers but it does not solve the main headaches of cleaning the raw data in real time and auto calibration functionality. EF EVE Volumetric Capture software is the number one destination for solving these two main problems at once. With a monthly subscription model it became well known for developers and all sizes of companies. It is one stop solution to start utilising Kinect Azure and Kinect v2 capabilities.
Audio matters
The Kinect Azure has an outrageous 7-mic circular array enabling extra layer of functionality.Few specs:
- Sensitivity: -22 dBFS (94 dB SPL, 1 kHz)
- Signal to noise ratio > 65 dB
The old Kinect v2 counts almost half of that with 4 mics.
Final word
It seems that Microsoft shifted from Kinect known as a gaming peripheral to a stand alone 3D camera leaving a door wide open to the new emerging technology of volumetric capture. Kinect now has 6 times better resolution, is twice as light and has better audio. It can be easily considered the best off the shelf sensor on the market for volumetric video. Now you can use EF EVE ™ automatic calibration with 2 or 4 Azure Kinect sensors and have your quality set up in less than 40 minutes.
