Colin Firth as My Celebrity Twin?

While facial recognition technology has been available for personal pictures on desktop computers for years (for example, MyHeritage allows you to look up your celebrity twin), Gizmodo just announced that the functionality is now migrating to the iPhone via an application called FaceDouble. Not sure I see the resemblance, but my Twitter profile picture came up at 73% Colin Firth.


Once again, Poncho was not recognized.

iPhoto Offers Face Rec for Cats...Not Dogs

Via Maclife:

A huge disappointment for pet owners, was word that iPhoto's newest feature,
Faces, wouldn't recognize animals according to Apple employees on the show
floor. We even did a few tests at Macworld Expo with images they had of the dog
Luce. Well, we are happy to say that we were able to get iPhoto to recognize
Robbie's Cat, Lola.

Poncho will not be happy.

Facial Recognition Training To Prevent Racial Bias?

(Credit: VizCogLab/University of Victoria)

Researchers from Brown University and the University of Victoria recently completed a study that suggests training people to recognize different facial features of individuals of a different race may reduce racial biases displayed unconsciously.

The landmark study involved 20 Caucasian participants, all of which were shown a series of black and white pictures of African-American faces in what was called an 'implicit association test.' Following each photo was either nonsense or a real word, of which the real ones had either a positive or negative connotation. The individuals responded immediately whether the word was real or nonsensical.

Initially, participants responded more quickly when a negative word followed the image and more slowly if the word was positive. However, in the same tests following 10 hours of training in which the first half learned how to distinguish among African-American faces and the other half learned to identify simply whether the faces were African American or not, individuals in the first group showed an increase in positive associations with the pictures and a decrease in negative ones.

"As soon as you can tell those people apart better and you can really tell that they're different individuals, then you'd be less likely to make an automatic generalization," said Michael Tarr, a professor of brain and cognitive science at Brown University.

Previous studies in this area have confirmed this 'learned aspect' and show that it is less difficult for participants to distinguish between faces of people in the same racial group in which they were raised. In fact, one project suggested that African-American children raised by white adoptive parents were more easily able to differentiate between white faces than black ones.

Although limited in scope, researchers would like to think this idea of recognition training could potentially be universally successful and signal a possible end to racial bias. Researchers are optimistic that these results could have implications in the real world, particularly for police officers, social workers and immigration officials looking to improve their differentiation of members of a racial group other than their own.

"The idea is this that this sort of perceptual training gives you a new tool to address the kinds of biases people show unconsciously and may not even be aware they have," said Tarr.

In fact, co-author Jim Tanaka believes the Obama administration taking office might be a perfect real world example of this study in action, which was actually released the day Obama was inaugurated.

He thinks that for the small segment of Caucasian folks with little interaction with those of African-American ethnicity, the tight media coverage of the First Family and the president's network will allow them to learn the features that make each person's face distinct and allow them to adapt this knowledge to the general population.

"I think clearly anytime you have positive examples to help break stereotypes is good," Tarr said. "President Obama does that."

While studies show that humans (and bees!) are born with innate, complex abilities to determine facial recognition, harnessing these skills to more precisely process features looks to be a developed talent. I'm thrilled to see the advances we're making in the research field and excited to see how they play out going forward. Perhaps seeing "Facial Recognition Training 101" in orientation programs for various public-facing career fields is not such a bizarre thought after all.

$17,000 Megapixel UAV-Spybot

Check out this video of new LP960 UAV via Engadget Spanish:

Nice music. The 10 mega-pixel imager in the current model, seems a little light to me, however, considering what is possible today.

Bees -- Good For More Than Just Honey

There's some interesting facial recognition research just out of Monash University in Australia -- and I'm not talking about research on human faces. Instead, scientists have proven that honeybees can distinguish between two faces. Yes, honeybees.

In the study, performed over two years in Australia and Germany, researchers trained different groups of bees to fly toward human faces printed on a vertical screen. Bees that made the correct choice were rewarded with sugar or a bitter solution for incorrect decisions, and over the course of a day or so, bees had learned the task.

The groundbreaking discovery came later when only the bees that had experienced multiple views of a face at both 0 and 60 degrees were able to correctly identify the same face at the rotational angle of 30 degrees in non-rewarded tasks. According to Dyer, this demonstrates a bee's tendency to solve rotational problems by either averaging previously-seen images or mentally rotating previously learnt views.

"Bee brains clearly use image interpolation to solve the problem. In other words, bees that had learnt what a particular face looked like from two different viewpoints could then recognize a novel view of this target face. However, bees that had only learnt a single view could not recognize novel views," Dr. Dyer said.

This is exciting news considering most artificial intelligence recognition systems have major difficulties in reliably recognizing faces from different viewpoints.

"What we have shown is that the bee brain, which contains less than one million neurons, is actually very good at learning to master complex tasks. Computer and imaging technology programmers who are working on solving complex visual recognition tasks using minimal hardware resources will find this research useful," he added.

Turns out honeybees may be useful for more than honey and flower pollination (and of course, their tremendous dance moves).

Update: Obama Photosynth Results

Well the results of the CNN/Microsoft collaboration are in. Take a look at what happens when you "synth" the collective photographs of an innauguration crowd into a single 3D model.





More on Microsoft Photosynth here.

DHS Privacy Workshop Delivers Best Practices

As sophisticated surveillance grids continue to pop up nationwide, it was only a matter of time before a group of experts convened to discuss the trajectory of Closed Circuit Television (CCTV) technology and explore best practices when installing the technology. While the Department of Homeland Security (DHS) hosted its "CCTV: Developing Privacy Best Practices Workshop" over a year ago, only recently did the executive report summarizing the results of the conference surface.

As the Executive Summary states, the report serves as a best practices guide to avoid crossing privacy boundaries when deploying of CCTV systems as systems become more pervasive in both the private and public spheres. However, rather than advocate installation, the report serves as an objective guide to the various concerns to consider when first pondering the massive investment involved. Attending academics, researchers and government officials deliberated the importance of ensuring safeguards and civil liberties prior to deployment:

"These resources may be useful in helping government agencies build privacy and civil liberties protections into the design and implementation of a CCTV program. Failure to address privacy and civil liberties can undermine public support for the use of CCTV and erode confidence in the government's ability to protect privacy and civil liberties while protecting the Homeland."

Six panels ranging from an in-depth look at CCTV technologies being used today to general international surveillance practices culminated in suggestions on how to comply with major concerns based on the Fair Information Practices Principles (FIPPs), a set of principles that have long served as a framework for protecting privacy within the United States and abroad since 1973.

With the Obama administration taking office earlier this week, we won't know the specifics on his agenda to "Protect Critical Infrastructure", but the new President's stated objectives do include improving airline security, monitoring US ports, safeguarding public transportation and improving border security.

With a more sophisticated surveillance blanket covering the US, a best practices guide is certainly necessary to preserve individual privacy and civil liberties. In following through on his agenda action items, Obama's cabinet might want a copy of this 66-page manual delivered to the White House sooner rather than later.

Men are from Mars, and Women are Green

The next big breakthrough in facial recognition might be closer to home than we imagined. According to an article posted on DailyTech this week, the key lies within the neural workings of the human brain. However, what remains to be seen is how exactly the brain processes distinct facial features to determine identity. Going forward, understanding this internal facial recognition functioning could lead to a whole new wave of facial recognition technology.

Facial recognition is a topic of great debate and research, and the article presents two main schools of thought: those who believe the human brain is hardwired to recognize faces and those who think it's something we learn, and there are numerous subsets within these two camps.

Some have focused on facial tones and found men's faces to be "redder" than those of women's, which are more of a greenish hue. Others believe that the nose is the first place we look to identify a face, while another group suggests that the eyebrows play central role in recognition. Another MIT researcher believes that we process faces through a specialized part of the brain devoted to recognition called the fusiform face area, while many other scientists wholeheartedly disagree that we're born with any face recognition hardware at all. A hot topic indeed.

As science continues to catch up with innate abilities, facial recognition technology continues to adapt and improve. Who would have thought that eyebrows might be the real windows to the soul?

Photosynthing the Inauguration

CNN and Microsoft are attempting to create a moment-to-moment high resolution 3D model of the Obama inauguration using only submitted photographs from excited Obama onlookers...and Microsoft's Photosynth software. The technology works by finding commonalities among hundreds, or thousands, or even tens of thousands, of individual photographs of a particular event or place. Once processed, each image's position in the broader 3D context can be calculated with its visual information becoming part of a larger model. Microsoft has already used the technology to create visually stunning models of the Sphinx, the Statue of Liberty and more.

From Microsoft Live Labs:

The world will change on January 20th when Barack Obama, the President-elect takes the oath of office and becomes the 44th President of the United States. "I do solemnly swear that I will faithfully execute the office of the President of the United States, and will to the best of my ability, preserve, protect, and defend the Constitution of the United States." If you are going to be among the millions attending, you can be part of history by helping create the most immersive and detailed experience of a single moment ever created.

If successful, in many ways the project will provide as detailed and comprehensive a picture of the occasion as any professional news broadcast. And, in the future there is no reason this technology couldn’t make leap to video, as well. Videosynth? We might be looking at the future of citizen journalism.

Biometrics Finding Its Way into Customs Checkpoints

It's already been an active new year for customs checkpoints.

Earlier this week, European Parliament approved the use of biometric data in EU passports, and yesterday the U.S. Department of Homeland Security (DHS) announced that upgraded biometric technology has been installed at every major port of entry, and that most visitors should expect to use the new technology upon entering the country.

The European Parliament will amend a 2004 regulation to include fingerprints and facial patterns in passports issued in the EU, thus making it more difficult for criminals to forge identification documents and/or travel under stolen passports. Additionally, the new EU regulations will set the minimum age to take fingerprints at twelve, as fingerprints change as children grow. Instead of being included in their parent's passports, all children will be required to carry their own in an effort to make child trafficking more difficult.

Back in the U.S., the DHS announced that it has updated biometric technology in its US-VISIT program, which records biographic information to conduct security checks and verify identities of international visitors to the United States. The program is also expanding the categories of non-U.S. citizens required to provide digital fingerprints and a photograph upon entry to the U.S. in December, and requiring 10 fingerprint collections rather than two. Many experts agree that collecting a full set increases matching accuracy and also reduces the chance of misidentification.

With increased wait times expected and obstacles associated with integrating the various Extended Access Control (EAC) systems now in use across the globe, it will take years for travelers and customs officials to get acclimated to the new procedures and documents. In the meantime, bring a magazine for that next trip through customs.

What’s Working and What’s Not in Facial Surveillance

Facial recognition technology has certainly been a hot topic in 2008 – and one that will likely only continue to make headlines as more companies and organizations recognize its unique value within their security infrastructures.

When most people think of facial surveillance, however, what often comes to mind is the technology’s very public failure at the Super Bowl and other large-scale public venues over the past several years. Though the technology has certainly improved, we are still not much closer to being able to spot a bad guy in a crowd than we were then, and this points to the fact that it is facial recognition’s role within the larger security platform that allows the technology to function most effectively, and ultimately most accurately.

Nonetheless, there have been some important facial rec successes over the past year that deserve mentioning – let’s take a look at what’s working, what’s not, and most importantly, why.

Identifying Suspects: Using Facial Rec to Compare Captured Images to Police Database Mugshots

The Sagem Morpho MorphoFace Investigate (MFI) system has scored its first arrest in Pierce County, Washington. The Pierce County Sheriff's Department used the MFI biometric facial recognition application to identify a suspect by comparing an automatic teller machine (ATM) photograph against the department's digital database of 350,000 mug shots.

In the past, the only way police could conduct these types of investigations was to endlessly flip through mugshot books -- this is a major improvement and compelling evidence of the next generation of crimefighting taking place.

Identifying a Suspect Using Commercial Surveillance and Transaction Records

Just the other month I learned about the following incident from a 3VR customer. A man had been followed home from a large African bank and subsequently robbed after making a large withdrawl. While the victim didn't recognize the thief, he was able to point him out in bank surveillance footage leaving the bank after the victim. Normally, this is where the investigation would have become difficult; actually identifying the suspect. However, in this instance, the bank was able to perform a facial search against it's own surveillance archives using its 3VR system. The thief, it turns out, was actually a bank customer. Even though he was not in any police database, they were able to identify him using bank transaction records. You see, this robber's tactic was to spend a lot of time in the bank doing small transactions and other petty business while he watched and waited for his victims to withdraw a large amount of money. But once even a single image was of him was captured, the robber's face led bank officials and law enforcement right back to those trasactions...and the theif's real identity.

Alerting Security When the Bad Guys Arrive

Though an uncontrolled venue such as the Super Bowl may provide too difficult a context to do real-time facial alerting, in more controlled venues like banks and some retail establishments and using relatively targeted top-quality watch lists, it is now possible to use facial alerting successfully.

Only two days after pilot installation of the 3VR platform at another large international bank, a person wanted for check fraud entered the bank accompanied by an accomplice, approached the teller and began a transaction. The 3VR system being utilized in the branch recognized the person and immediately sent an alert to the bank’s security personnel, who compared the image to photos in order to confirm that it was indeed the suspected fraudster.

The bank’s security personnel were able to quickly contact the police, apprehend the woman at the branch and question her. She ultimately admitted to the fraud — case closed.

Looking Ahead

Facial surveillance has developed significantly with new technology and new approaches making up for many past failures. And while still certainly not perfect, modern "facial surveillance" represents a quantum leap forward from they days when all police had to go on were “WANTED” posters, mug shot books, and their own eyes and energy.

Getting San Fran Surveillance Right

Surveillance cameras mounted at First and Mission Streets in San Francisco

Big news. Despite doing many things wrong, San Francisco has still managed to reduce non-violent crime by an average of 24% in areas where they deployed surveillance cameras. Yet, a main objective to reduce violent crimes has yet to be accomplished. But, what would it mean if the city did everything right? That question wasn't addressed in a substantial piece in the SF Chronicle today, but the statistics the article provides are making for some pretty interesting food for thought.

The article cites a recently released study that shows that San Francisco's surveillance program has failed in its primary goal of reducing homicide and other violent crime, although it has succeeded in reducing such lesser offenses as burglary, pickpocketing and purse-snatching. While this statistic is important, it should be noted that San Francisco's cameras are not monitored in real time, but rather the footage is ordered by investigators only after the crime is reported; privacy controls prevent city police from monitoring the city's surveillance cameras in real-time. The privacy controls are so stringent, in fact, that I recently wrote about a man forced to spend 69 days in a San Francisco jail waiting for access to surveillance video footage that ultimately exonerated him. The city can and should be doing much better.

Protecting Privacy With Technology vs. Non-Use

There is very little excuse these days for deploying city surveillance and then not monitoring cameras or granting police and attorneys access to footage that helps convict or exonerate. That's because identity protection and search engine technology exists today that enables police to actively monitor and query city surveillance systems without sacrificing privacy. The ACLU recently wrote about one example of this kind of technology from 3VR Security, and these kinds of approaches are getting better every day. Had San Francisco deployed its surveillance system with this kind of privacy protection technology, it would not have been necessary for the city to take the extreme step of locking police out of its surveillance infrastructure.

The Value of Image Quality and Data Storage

In San Francisco, image quality and data storage are two other major concerns. The cameras the city purchased and installed for $700,000 in 2005 are high-resolution, but produce only three frames per second and thus the footage appears choppy (movies are shot at 24 frames per second), making it difficult to identify even license plates – forget about trying to recognize a repeat car thief or vandal. Again technology may provide an answer for the city. Instead of storing everything and an equally poor frame rate and video compression level, why not use face and license plate detection algorithms to smartly track and store relevant information when, and only when, it is detected. Intelligent approaches that separate what is important from what is not have been proven to dramatically improve both the quality and storage longevity of surveillance archives.

A Fully Integrated Platform

Traditional surveillance systems often require an army of people to patrol the system and report suspicious activity to a security manager. However, by using networked systems that apply analytics and reporting, municipalities can service security functions of hundreds with an army of one. By incorporating the latest innovations in face recognition, license plate recognition, video motion alarms and other new technologies to make systems more effective and efficient, municipalities can vastly improve the results of their security systems. As John Honovich points out in a recent contribution to Government Security News, "cameras enable officers to assess and respond over much greater areas at much lower cost. Even in the U.K., famous for its mass public deployment of surveillance cameras, video surveillance costs are only one one-hundredth of police costs."

Creating A Better Solution

Ultimately, video surveillance is nothing new and many cities are capturing hours and hours of footage on a daily basis. However, that endless volume of footage must be managed to be valuable, and even if a surveillance system features the highest quality cameras, the best image resolution, and a fully integrated network, the resulting video footage isn't going to be useful unless it can be stored, analyzed and searched – in real time. Search and analytics would dramatically improve the usability of the footage and make the surveillance network that much more effective in achieving its objective of crime reduction.

With the right ingredients, municipal surveillance has the potential to significantly counter criminal activity, but it takes that correct formula and a scientific approach to have such results. Despite its inefficiencies, the San Francisco municipal surveillance system isn't a disaster by any means. If Newsom can bolster its allotment of the 2009 budget and add the aforementioned features into the existing infrastructure, San Francisco has the opportunity to have one of the largest municipal surveillance systems out there operating smoothly, meeting and perhaps even surpassing expectations.

Obama Onboard for Biometrics Spending

As we reported last month, biometrics spending worldwide is on the rise, and it looks like our new president won't be straying from the trend.

According to a recent study, President-elect Barack Obama and his administration team won't skimp when it comes to biometrics spending during their first year in office. The Stanford Group Co. research expects Obama's team to spend up to $1 billion on biometric applications, primarily in defense, intelligence and homeland security sectors. An expected $500-600 million will go to biometrics contracts, many of which were announced Tuesday, and additional intelligence programs may add another $250-350 million in expenditures.

A billion on biometrics? That's no small figure, and we look forward to tracking its development and impact in the years ahead.

Furry Thief Attempts Five-Fingered (or should we say Four-Legged) Discount

He's not your typical bank robber: