4 ms·

Yeah... The most worrisome aspect of the presentation for me was that it's not designed to work outdoors. They said explicitly that the initial product is aime

by shawn 8y ago

Yeah... The most worrisome aspect of the presentation for me was that it's not designed to work outdoors. They said explicitly that the initial product is aimed at indoor use. Which makes me wonder what it has that my Vive doesn't, since the field of view is apparently limited. (One of their presentations in July is titled "FOV mitigation.") Still, Magic Leap has proven time and again that they have a plan, and mixed reality is coming one way or the other. ML won't be the last player in the space.

baxtr 8y ago
Well, the whole thing reminds me of the Segway story back in the days. That didn’t play out as expected
- kuschku 8y ago
  If this goes like Segway, then we can expect 10 years later cheap chinese clones, slightly modified to show up and become popular in every household. As happened with the "hoverboards" that are basically just cheap clones of the SegWay concept.
  - ghaff 8y ago
    "Becomes popular" is more like "becomes a short-lived fad." No? I don't have sales numbers but there are any number of articles suggesting that they came and went. [1] Certainly, one doesn't hear or see much about them these days. [1] http://fortune.com/hoverboard-industry/ http://fortune.com/hoverboard-industry/
    - Larrikin 8y ago
      I think they wouldn't have become a fad if they were well made and didn't become huge fire hazards that were disallowed on airplanes. The decline in public use I saw was tied to the number of self combusting articles in the news.
      - ghaff 8y ago
        But then they'd probably have been a lot more expensive and less interesting as a result--back to the upstream point. I actually think there were broader issues with Segways; they don't really work on either sidewalks or roads. But price didn't help.
joshvm 8y ago
The main differences with indoors vs outdoors are illumination and distances. In terms of motion tracking, outdoors tends to be a lot easier because there are far more varied/unique textures than indoors. Outdoor environments get a huge amount of NIR light which can swamp cameras without an IR cut filter. It's possible that the Leap is using infrared imagery for motion/depth tracking. Similarly devices which project structured light for 3D reconstruction usually use NIR projection. These devices fail in sunlight. The other issue is depth. Depth sensing on a device like the Leap is going to be done using either time of flight, structured light or stereo [1]. Time of flight is unlikely due to the amount of power you need to dissipate on the LEDs (see the Kinect V2) - which are mounted next to your face. This approach would also preclude outdoor use, see above. If they're using stereo alone, then the depth accuracy will only be good at close range. For a typical baseline of say 20 cm, you're going to get reasonable results up to 5 metres [2]. More robustness if you use structured light to assist. [1] There appear to be multiple cameras, so why would you not exploit the geometry? [2] Back of the envelope calculation: stereo error at 10m using a 20cm baseline, 2.2um pixels, 5mm focal length and 0.5px accuracy is about 5cm. Accuracy at 5m is about 3cm.
- AstralStorm 8y ago
  You could get a bit better than stereoscopy by using say 4 cameras. Plus since it is on moving head, you could use a bit of surface from motion techniques. You could also make the cameras rotate on their own (or mirror - like DLP; or lens), like eyes do, to enhance the result when head does not move. Loads of processing power required... solvable. Now the real problem is the display of sufficient brightness and colours. Some sort of future translucent LED screen could do it, maybe. DLP can do it at big cost and with bulky projectors. Current picoprojectors are not very bright... And then three is a focal length problem to solve to have sharp image and good depth range while not messing with accommodation.

by shawn 8y ago

baxtr 8y ago

Well, the whole thing reminds me of the Segway story back in the days. That didn’t play out as expected

kuschku 8y ago
If this goes like Segway, then we can expect 10 years later cheap chinese clones, slightly modified to show up and become popular in every household. As happened with the "hoverboards" that are basically just cheap clones of the SegWay concept.
- ghaff 8y ago
  "Becomes popular" is more like "becomes a short-lived fad." No? I don't have sales numbers but there are any number of articles suggesting that they came and went. [1] Certainly, one doesn't hear or see much about them these days. [1] http://fortune.com/hoverboard-industry/ http://fortune.com/hoverboard-industry/
  - Larrikin 8y ago
    I think they wouldn't have become a fad if they were well made and didn't become huge fire hazards that were disallowed on airplanes. The decline in public use I saw was tied to the number of self combusting articles in the news.
    - ghaff 8y ago
      But then they'd probably have been a lot more expensive and less interesting as a result--back to the upstream point. I actually think there were broader issues with Segways; they don't really work on either sidewalks or roads. But price didn't help.

joshvm 8y ago

The main differences with indoors vs outdoors are illumination and distances. In terms of motion tracking, outdoors tends to be a lot easier because there are far more varied/unique textures than indoors. Outdoor environments get a huge amount of NIR light which can swamp cameras without an IR cut filter. It's possible that the Leap is using infrared imagery for motion/depth tracking. Similarly devices which project structured light for 3D reconstruction usually use NIR projection. These devices fail in sunlight. The other issue is depth. Depth sensing on a device like the Leap is going to be done using either time of flight, structured light or stereo [1]. Time of flight is unlikely due to the amount of power you need to dissipate on the LEDs (see the Kinect V2) - which are mounted next to your face. This approach would also preclude outdoor use, see above. If they're using stereo alone, then the depth accuracy will only be good at close range. For a typical baseline of say 20 cm, you're going to get reasonable results up to 5 metres [2]. More robustness if you use structured light to assist. [1] There appear to be multiple cameras, so why would you not exploit the geometry? [2] Back of the envelope calculation: stereo error at 10m using a 20cm baseline, 2.2um pixels, 5mm focal length and 0.5px accuracy is about 5cm. Accuracy at 5m is about 3cm.

AstralStorm 8y ago
You could get a bit better than stereoscopy by using say 4 cameras. Plus since it is on moving head, you could use a bit of surface from motion techniques. You could also make the cameras rotate on their own (or mirror - like DLP; or lens), like eyes do, to enhance the result when head does not move. Loads of processing power required... solvable. Now the real problem is the display of sufficient brightness and colours. Some sort of future translucent LED screen could do it, maybe. DLP can do it at big cost and with bulky projectors. Current picoprojectors are not very bright... And then three is a focal length problem to solve to have sharp image and good depth range while not messing with accommodation.