I’ve been lucky enough to have tried the Oculus Rift DK2 on numerous occasions prior to its release this week, but never for a long enough that I felt like I really knew the device. After receiving the DK2 this week, and finally able to spend as much time as I want investigating its performance, I’ve come away beyond impressed with its capabilities, particularly its positional tracking prowess.

One of the Oculus Rift DK2’s biggest improvements over the DK1 is its ability to positionally track the headseat. The DK1 tracked rotation very well, but if you were to lean at all with your torso, your virtual head would not move accordingly. This disconnect was a big issue for the DK1 for both immersion and sim-sickness. Even without purposefully leaning in to try to inspect something in the virtual world, we make lots of subtle movements—like shifting in a chair—which can’t be captured with rotation alone.

ces 2014 oculus rift crystal cove prototype interview palmer luckey nate mitchell low persistence positional tracking
The Oculus Rift ‘Crystal Cove’ prototype left the IR-LED array exposed while the DK2 covers the array in an IR-transparent housing.

To solve this, Oculus VR opted to build an IR-LED array into the DK2. The array is tracked by an included CMOS camera which captures frames 60 times per second. Oculus have done an outstanding job with their strategic placement of the array and with the computer vision work that allows the DK2 to handle occlusions with aplomb. For a single camera ‘outside-in’ setup, this is definitely the best tracking performance (in terms of capture size, quality, and ease of use) that I’ve ever seen at this price.

I wanted to see how well the DK2’s camera could handle distances and motions beyond what your typical seated scenario demands… turns out it can handle much, much more.

Newsletter graphic

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. More information.

Ben is the world's most senior professional analyst solely dedicated to the XR industry, having founded Road to VR in 2011—a year before the Oculus Kickstarter sparked a resurgence that led to the modern XR landscape. He has authored more than 3,000 articles chronicling the evolution of the XR industry over more than a decade. With that unique perspective, Ben has been consistently recognized as one of the most influential voices in XR, giving keynotes and joining panel and podcast discussions at key industry events. He is a self-described "journalist and analyst, not evangelist."
  • Stimpack

    This is the kind of content I love to see. Very exciting times for VR. Cannot wait to see what the years will bring us, and this is definitely major leaps in the right direction.

  • eyeandeye

    Fantastic post. Have you tested the maximum working distance from the camera? I love the part where you lay down on the floor. I will definitely need to find a camera position that will allow me to do that once I get mine. Nice to know that it will be easy to play with different camera locations.

    I keep thinking about how mean a prank it would be to walk in on someone playing in the DK2, quietly snatch their camera and wave it around.

    • eyeandeye

      I also meant to ask if you ever felt dizzy or disconnected while standing up. Even when trying to stand perfectly still in DK1, I could only keep my balance if I had my legs touching the edge of my desk or chair.

  • spyro

    The problem: Where to place your camera? Oculus says, it should be placed in the same height as your eyes, 1.5 m away. Most people sit about 40-50 cm away from their displays. So if you mount it there, it far to close (your head will constantly be out of the camera’s view frustum). Behind the display is a wall in most cases. If you simply slide away 1 m from your desk, you can’t reach your keyboard and mouse (which are unfortunately needed for most applications). So why didn’t they use a much more wider angle lens? Maybe it’s possible to mod this?

    • Simon

      Great video!

      @spyro: Since the camera is using just using the viewed image, I wonder whether you could use a couple of mirrors to fold the view (maybe like a periscope) to allow the camera to be physically closer, but still maintain the recomended distance.

    • mptp

      This is really great – awesome to see the effect of different levels of occlusion on the tracking accuracy. Thanks so much for this!
      I personally think that the inability to turn more than 180˚ is a serious problem. I get that it’s not easy to fix when using a single camera, and being able to achieve 180˚ rotation with a single camera is a huge achievement by OVR, but that doesn’t change the fact that it’s a big problem. The first thing anyone does when they get into VR for the first time is look all around them, especially behind. I don’t think I’ve ever seen someone not try to look straight back within the first 30 seconds of their first VR experience. It’s one of the important things that VR can do that a 2D screen can’t do.
      It wouldn’t be such a problem in traditional first-person experiences since you rarely look behind you when walking forwards, but for the seated, cockpit-style experiences that OVR is endorsing so strongly, it’s going to be a common gesture. The fact that the Rift just flat out drops positional tracking when performing this motion isn’t going to cut it for CV1.
      So I wonder what Oculus is going to do to try and offset this issue? Another camera? Inside-out tracking? More IR points?
      I just hope they do something. ;)

      I couldn’t find the forum thread, but there was a guy on the OVR forums who jerry-rigged a fisheye lens onto his DK2.
      While it does increase the tracking volume, it also distorts the locations of the markers on the camera frame dramatically. So he would move his head forward like 20cm, and his virtual head would move forward two virtual metres.
      It could be fixed at the software level, but it wouldn’t be a trivial fix.

      I think the fact that the camera needs to be placed 1.5m away isn’t a huge deal, it’s inconvenient for us now during this transitional period, but soon everyone will be considering the placement of tracking cameras when deciding where to put their desk, and it will be less of an issue.

      Hmm…I’m having difficulty visualising how you could extend the viewing angle of the camera this way without distorting the position of the IR markers. If it would distort the IR marker position, you’d have the same problem as fitting a wide-angle lens to the DK2 camera – that is to say, solvable at the software level, but non-trivial.

  • sponge101

    This vid makes me feel much better buying an Omni or Virtulizer. Thanks.

  • SimplyKen

    “The best tracking performance at this price”? Uh. There are literally ZERO VR HMDs on sale right now. What are you comparing it to?

    • Wmerr21

      I think he is referring to 3rd party tracking hardware for the oculus dev kit, such as STEM system and PrioVR.
      I would be interested to know how the native optical positional tracking compares to these other systems in terms of accuracy and latency, as it seems like you would need about three cameras to achieve 360 degree tracking this way.

    • Jarom Madsen

      The Kinect as mentioned, wii remotes, sixense hardware (hydra/stem), any other motion capture systems, etc. For low-latency tracking vs price, it’s a good piece of kit.