Development of Real Time Face Recognition System using OpenCV
Tech Review Genl
1. ANALYSIS OF TECHNOLOGIES CURRENTLY
AVAILABLE TO CREATE PHOTO REALISTIC
3D AVATARS
By Olivier P. Sarda and Eric P. Ashenberg
Los Angeles, March 5, 2010
INTRODUCTION
As readers of the Simetrix Imaging, Inc. (“Simetrix”) Business Plan are aware, it is
the Company’s vision to become the world’s first entity to create, maintain and
protect a database of 3D photo-realistic consumer avatars to be used in a host of
commercial applications. Accordingly, the Company has extensively researched the
3D capturing technologies currently on the market and in-development to
determine which method(s) would best help achieve its goals.
Numerous factors were considered in making this review including: cost, ease of
implementation, quality of images, reliability, scalability, ease of use for repurposing
of images for other applications and many others. Though some factors may seem to
be diametrically opposed to another, all were duly considered and analyzed.
SUMMARY OF CONCLUSIONS
Simetrix believes that currently, only a standardized photo capture system can
provide the precise images necessary to create high quality photo realistic 3D
avatars of consumers that can be leveraged into a host of retail applications. A
standardized system using uniform camera and lighting equipment will best assure
that all images captured will be reliable and appealing to showcase merchandise to
consumers. The Company will continually monitor innovations in 3D capturing
technology and will be prepared to switch to a user generated photo system as soon
as such systems yield images of sufficient quality capable of producing fully accurate
3D avatars.
2. 2
PROFESSIONALKIOSK PHOTOCAPTURE STATION
(Simetrix Photo Capture Station Design Concepts) (Photo-real Wireframe/ Avatar)
How it works: Simultaneous multiple photo capture system with ability to
extrapolate photo data for realistic 3D avatar creation.
This is currently the Company’s preferred choice of technology simply because of its
high-quality output and life-like/photo-real 3D avatar creation capabilities.
o Pros
o Photo-realistic 3D avatar creation
o Super high-resolution (scalable and dependent on continuous
advancement of D-SLR camera or other camera technology, e.g.
Z-Depth or cyberscanners)
o Controlled & consistent lighting / angles / resolution
o Top down technology / highly scalable for ecommerce
o Ability to downscale avatars for all kinds of use (Smartphones,
Internet, Press, etc…)
o No user interaction needed to create avatar
o No software to install
o No webcam needed
o More future proof / less obsolescence built-in
o Cons
o Higher initial upfront costs
o Hardware/software support necessary
o Need basic training to operate
o Requires full scale marketing campaign to launch
3. 3
ONLINE USER GENERATEDAVATARS BASED ON PHOTOS
(Uploaded User Photos) (Defined Markers Placed On Facial Attributes by User)
(Avatar Result From Photos)
How it works: Users upload one or more pictures of themselves (usally 3 photos -
one fontal view and right and left side profiles). Requires users to manually confirm
location of facial attributes by placing points/markers on photos.
o Pros
o Extremely affordable method
o All software driven
o Low upfront costs
o No hardware support necessary
o Webcams widely available around the world
o Cons
o No quality control often results in low resolution avatar
o Avatar is an approximation of user facial characteristics (usually a
look-alike representation but not believable recreation of one’s
face)
o Inconsistent results from poor lighting, unflattering angles and
low camera resolution often results in a “plastic” image
o Can’t usually show hair but rather a conformed bald head
o Does not provide a precise virtual fit
4. 4
o Bottom up technology / not scalable
o Involves user interaction without prior knowledge of individual
user’s experience and demographic profile
o Limited by requiring users to place markers on facial points before
creating 3D image
AUGMENTEDREALITY
How it works: Live video feed with graphics inlaid on top in real-time; in essence,
the real world is augmented by virtual computer generated imagery.
o Pros
o Amusing results on actual live video feed
o Fun & interactive when working properly
o Cons
o No proper virtual fit
o Does not track properly / glasses float on face
o Low resolution mostly based on limits of user webcams
o Limited to front of face / cannot do profiles or simply very
slow head movements from left to right as shown on photo
above
o Need to install proprietary software to work
o Involves extra user interaction without prior knowledge of
individual experience and demographics
o Feels gimmicky
5. 5
INTERACTIVE ONLINE USER PHOTOS
How it works: Photoshop style overlay of glasses on uploaded user photo taken
from a user webcam or photo library.
o Pros
o Simple to use
o Quick webcam low-resolution photo capture
o Fast upload of user photos
o No software install needed
o Cons
o No virtual fit
o 2D approximation with no real world precision
o Extremely gimmicky for eCommerce use
o Unable to represent actual scale ratio of product to face
(glasses fit a baby the same way it would an adult)
o No motion / completely static
o Low resolution mostly based on limits of user webcams
6. 6
Rapid Model AcquisitionOnline
How it works: This technology is capable of turning any standard webcam into a
powerful 3D scanning tool.
o Pros
o 3D models constructed in real-time as end-users slowly rotate
objects with their hand in front of webcam
o Great point tracking technology from motion estimation
o Could potentially be adapted to capture faces in 3D as user
rotates head
o Could provide better results in the future as webcam
resolution increases
o Cons
o Not adaptable for human avatars or curves of human face –
meant more for simple geometric shapes
o Technology still in early stages
o Limited by user webcams
o End result is still problematic and inconsistent in quality
o Only constructs partial geometry
o Lighting & resolution are affected
CONCLUSIONS
While Rapid Model Acquisition Online technology holds much future promise,
currently it does not have the ability to fully capture the complex geometry,
curvatures and angles comprising the human face. Only a professional kiosk photo
capture station provides a uniform, high quality solution to creating reliable and
precise avatars that can continually be used to showcase a range of consumer
products without fear of obsolescence. The Company will actively monitor
developments and innovations of all 3D imaging technologies.