Video Recognition

Recognition performed by humans

What is Video Recognition?

For humans, video [data] is a source of information about objects and activities captured by a video-camera. This is because human brain is able to perform many Video Recognition tasks.

For computers, video [data] is nothing but a collection of changing in time numbers (colour intensities)... unless a computer is able to perform Video Recognition tasks.

Recognition performed by computers

"Video Recognition" is a term that was introduced by Dr. Gorodnichy in to name a new area of science and technology that deals with recognition of objects and activities in video (Refs. www.videorecognition.com and www.computer-vision.org/VideoRec07).

Read about another term introduced by Dr. Gorodnichy: "Face Processing in Video".

Other definitions of this area include: Intelligent Video, Video Analytics, Video Analysis and Content Extraction, Perceptual Vision. It also related to recently established technology known as Smart Cameras.

The applications of automated Video Recognition are enormous and highly demanded. - If you have a TV, VCR, web-cam, DVD, or any other video playing or storing device (which you do), you likely need it too!

Examples of Video Recognition Tasks

Examples of Video Recognition Applications

Objects detection, tracking and recognition (inc. foreground and background)

Scene recognition (detection and annotation)

Activity, Event, Action detection and recognition

Video - audio processing

People tracking:

public surveillance,

information security,

traffic monitoring

biometrics (esp. "at a distance" or "on the move"),

industry for disables

computer-human interaction and hands-free control

vision interfaces (Nouse) and multi-media,

computer games, entertainment,

immersive and collaborative environments,

tele-conferences,

video coding,

video annotation,

digital television

intelligent video processing, video quality improvement

Real-time tracking,

backtracking ,

Multiple-people tracking,

Multi-camera tracking

Video recognition tasks dealing with processing of faces:

face segmentation

face detection: in image and 3D world space

face tracking: in image (2D) and in space (3D)

face localization (precise tracking)

face classification

face identification (often also referred to as face recognition)

face annotation

face memorization

face registration, acquisition, and archival

Many of these applications deal with recognition of faces, where by "recognition" we imply all types of video recognition tasks dealing with faces (as listed at left).

All of these tasks and applications have been dealt with in work of Dr. Gorodnichy.

-

Video data vs. Still imagery: Natural trade-off between temporal vs. spatial resolution

What evades attention of most Video Technology users and developers is the fact that video data is critically different from still imagery - it is, by nature, of high temporal resolution and low spatial resolution (e.g 1000 frames, 320x240 each showing a face), whereas still imagery is the opposite (eg. one 1600x1200 picture of a face). Hence the algorithms developed for recognition in still imagery, of which there have been many developed over the last several decades, may not be simply transferred to video domain.

This is what makes finding solutions in Video Recognition very challenging and interesting at the same time - the unique niche it occupies in research space. - Not only the traditional expertise inImage Processing and Computer Vision is required, but also that in Machine Learning & Pattern Recognition, as well as in Neurobiology and Artificial Neural Networks.

Real vs. fake intelligence: easy to test

What also makes research in Video Recognition even more fascinating is the affordability and accessibility of it. Everybody has access to video of some sort now. - Web-cams, handheld cameras, PDA's, internet-streamed videos and archived digitized videos provide abundance of media to be analyzed. This makes testing and demonstrating the value of Video Recognition technology easy -If your Video Recognition software is as intelligent as you claim, just make it available for others to try!

This is what Dr. Gorodnichy did - instead of just showing "nice" snap-shots or pre-recorded demonstrations, which is what most "Intelligent Video" companies still do, he made his prototypes freely available for public download and testing. And the word spread around - by 2003 his work is mentioned in "Recent Advances in Computer Vision" along with major Computer Vision Labs of USA, by 2004 it is rated over such computer giants as IBM, Microsoft, Apple, HP by a Brazilian review journal "Planeta digital", and by 2006 his technical pages have been visited by over a half a million visitors worldwide.

-

"Video Recognition Systems" (VRS) project, NRC, Insitute for Information Technology (2001-2007)

"At right time at right place" - that's how Dr. Gorodnichy felt back in 2001 when, having a freedom of choosing a new research direction within a newly formed Computational Video group of the National Research Council of Canada (NRC) and realizing that the "Era of real-time video processing" has just arrived and searching for the best fit for his two-sided expertise (PhD on Neural Intelligence and PhD on Robot Vision), he established a new project focused on real-time processing and recognition of video data.

This project, first called Perceptual Vision (www.perceptual-vision.com) and then renamed to Video Recognition Systems (www.videorecognition.com), became the first in Canada and one of the first in the world with the focus on developing technologies for real-time automated understanding of video data.

This was an act of visionary. - Dr. Gorodnichy is quoted from his paper published at the IEEE Face and Gesture Recognition conference in 2002: "Our presentiment is that soon most laptops will be equipped with build-in ‘eye’ (camera) above the screen" and five year later indeed such laptops appeared in the stores.

Visit www.videorecognition.com for more details on technologies and knowledgbase developed by within the VRS project!

-

CBSA Video Surveillance and Biometrics (VSB)Applied R&D Section (2008-2013)

Following the termination of the Video Recognition Systems project by NRC in November 2007 due to Work-Force Alignment, Dr. Dmitry Gorodnichy played a critical role in founding the new section within Canada Border Services Agency's Science & Engineering Directorate, where the NRC Video Recognition expertise and knowledgebase related to the needs of the agency is transferred to. The new section, named Video Surveillance and Biometrics section (VSB) has been created in January 2009 to become the first within the GoC to provide highest-level expert support in both Biometrics and Video Analytics.

-

IVIM Inc. - "Intelligent Vision Interfaces and more..." (2007 - present)

Visit www.IVIM.com and www.Nouse.comfor more details on Nouse technology and services offered by IVIM Inc.!

At the same time, to safeguard the NRC Video Recognition expertise and knowledgebase that is not related to the border services needs of the agency, such as Nouse® Hands-free Interface, which has been relied upon by many people worldwide and, in particular, by Ottawa Long-Care facilities, in November 2007, Dr. Dmitry Gorodnichy and his wife Alexandra Gorodnichy (MBA) incorporated a company IVIM Inc., to whom Nouse® has been transferred. See more: at www.Nouse.ca

Looking for an advise in Video Technology or have a business proposal ? - Don't hesitate to contact Dr. Gorodnichy by email: dg @ ivim.ca = dg @ videorecognition.com = dmitry @ gorodnichy.ca.

Want to try our Video Recognition technology ? - Some of it available for public for free evaluation. Download Nouse® Hands-free Interface from IVIM's website: www.ivim.ca !

See also the publications and talks Dr. Gorodnichy at www.videorecognition.com/doc. and the knowledgebase and services he created at www.computer-vision.org

References: