A tidal wave of new wearable tech, 3D sensors and displays is coming that will bring computers into our 3D world, and it's coming this year or the next. This presentation goes over the enabling technology (depth sensor, object tracking & recognition algorithms, better cpus and batteries), surveys several new devices coming out, and finally talks about the societal implications of having computers so much more tightly integrated into our world.
7. Sentrinsic Background
● Formed in 2006 on patented linear sensing tech from
GaTech lab
● Developed embedded control
system to improve efficiency of
Air-Operated Diaphragm Pumps
8. Sentrinsic Background● Mobile Integration
○ Bluetooth Pump Monitoring
○ Customer iPhone app
■ Scan pumps for documentation, order
parts, track maintenance
■ Product: EquipCodes.com
● Augmented Reality
○ Maintenance
○ Operation Theory
○ Training
● Google Glass Manufacturing
Assembly Training app EquipCodes app
9. Augmented Reality
● What is it?
● Examples
○ 1st & 10 yellow line
○ Simple text
■ translations
■ nearby attractions
■ instructions
14. Texture Feature Matching
● Find interesting points (corners)
○ independent of orientation, scale, lighting
● Use surrounding area to develop fingerprint of feature
(128 pt vector, robust to skew).
● Based on neuron
receptors in visual
cortex, sensitive
to spatial gradients
Distinctive Image Features from Scale-
Invariant Keypoints, Lowe 2004
15. Also used for Image Stitching - Using Feature Recognition
17. Texture Feature Matching
● Compare feature vector against database (Euclidian
distance).
● Can’t be regular pattern (checker board)
● High contrast required
● Glare problematic
● Tracker must be large
relative to graphics to
prevent jitter, misalignment.
18. Extensible Tracking
● Environment unknown beforehand
● SLAM (Simultaneous Localization and Mapping) or PTAM (Parallel
Tracking and Mapping)
○ Structure from Motion - Autodesk’s 123D
○ Scene should be static
○ For overlays, requires initial registration, not for games
○ Videos: http://www.youtube.com/watch?v=i7ierVkXYa8 (3D structure from 2D pictures)
○ http://www.youtube.com/watch?v=Y9HMn6bd-v8#t=50 (real-time map generation on iPhone)
● For overlays that must match exactly, still need some initial alignment.
○ Edge Based tracking using 3D Models
■ Uses extensible tracking, initialized with best
fit edge of model. (Metaio SDK, Sept, 2013)
○ Small tracker can still be used
20. Depth Sensors
● Skips over many challenges with regular
camera tracking
○ Directly measures object depth
○ Can handle “featureless” textures (everything same color)
○ Still requires object recognition
○ Primarily used for body / hand tracking
● Examples
○ Kinect, Leap Motion, Laser range finder
21. How Kinect (ver 1) works
● Structured Lighting (Infrared)
video: http://www.youtube.com/watch?v=uq9SEJxZiUg
22. Kinect (ver 1) Drawbacks
● Affected by Sunlight
● Multiple kinects would conflict
● Minimum distance several feet - bad for
robotics and hand tracking (low resolution)
● Very cheap, though!
23. Time of Flight Sensor
● Uses speed of light and arrival time of reflected pulses
to calculate distance
● Extremely small times:
○ D = t * c / 2, 1cm → 66ps
○ 3 GHz cpu → 333ps
● Advantages: more compact (no separation between
sensor and emitter), simplier calculation, higher
resolution
● Kinect v2, Soft Kinetic sensor.
24. 3D Sensing Applications
● Augmented Reality
● Body tracking
○ general computer gestural control, more natural
○ gaming - Dance games
○ head tracking - CastAR, Virtual Reality
○ 3D design / printing
25. Google Project Tango
● Adding depth sensors to phones
● Incorporating mapping and tracking
technology
● Applications
○ visually impaired navigation
○ instant measurement of
home dimensions
○ store & shelf navigation
video: http://www.youtube.com/watch?v=Qe10ExwzCqk
26. Wearable Phones?
Many of the applications make
more sense if you don’t have to
hold a phone!
● Maintenance - hard to hold
iPad with wrench!
● Mapping
27. Wearable / Pervasive Computing
● Google Glass
○ Bad for AR?
■ Battery life
■ Camera can’t be on constantly
■ Small field of View, not in direct
line of sight
○ Can’t focus on glass and outside
world - looking on wiki mid conversation
30. Google Glass Pros
● We check our phones 110 times / day
● Better for contextual information
○ Incoming emails and texts
○ nearby places of interest
○ Instructions - following and creating (mfr DEMO)
● Voice recognition is great
● Capturing spontaneous or 1st person videos
● Looks cool, lightweight, viable wearable
● No tether
31. Head Mounted Display History
Steve Mann
Thad Starner - GaTech!
from 1993
Wearable Tech not new! Just
more compact and marketed!
Battery & CPU finally getting
small and cheap enough.
32. More head mounted options
● GlassUP
○ $300
○ small screen
○ tethered
33. ● Recon Jet
designed for Sporting
$600, shipping May?
● Meta Pro
○ Does what people think Google Glass does
○ built-in depth sensor, potential
Iron Man / Minority report int.
○ 40 degree Field of View
(vs 14 degree on Glass)
○ $3650, shipping Sept
○ tethered controller
34. CastAR & Spacial AR
● Projection on reflective mat
● Head tracking via LEDs
● Hologram effect
● form of Spatial AR
interaction
w/ projections
35. Oculus Rift / Crystal Cove
● Primarily for gaming, Virtual Reality
● Fully immersive 110 deg Field of View
● Head tracking via accel, gyro sensors,
markers on ver2
● Some have added cameras to achieve
Augmented Reality effect
○ Solves delay problems with see-thru displays
● Motion sickness problem - display must
react quickly to head movement.
● Requires fast frame rates to trick brain
Video: Oculus Rift with 2 Cameras
achieves Augmented Reality
http://www.youtube.com/watch?
v=Bc_TCLoH2CA
36. Technology Hurdles
● Latency: see-through AR must be faster.
○ Google glass small screen prevents motion sickness,
but lag will be more jarring in large screens
● Dimming screen
○ per-pixel dimming not viable yet, see-through text can
get washed out. (Video pass-thru helps)
● Better, Smaller, Faster (Battery, CPU, screen)
● Object & Gestural recognition algorithms
37. Societal Implications
● Always on camera
● How do we have a conversation with computer
distractions? Wearable manners?
● Engineering - Design in 3D with hands, more natural
gestural control
● Job Creation?
○ Reverse of robotic automation - use adept human dexterity
○ Enhance human capabilities via internet, AI or remote assistance.
○ Remote diagnostics, field service, DIY repair
38. Conclusion & Final Thoughts
● Consumer Computers sense and project in
3D space
● Technology more pervasive / always on via
wearable implementations
● Content is major bottleneck
● If useful, appearance won’t matter
39. Contact
● Scott Driscoll
○ Scott.Driscoll@sentrinsic.com
○ plus.google.com/+ScottDriscoll
○ blog: ImponderableThings.com
○ youtube: CuriousInventor
● Company Website: EquipCodes.com
● Software Tools
○ Vuforia, Unity, Metaio, Google App Engine