Was quite intrigued by the below new item. Sounds like tech going too far? How do you feel about running to the nearest 7 eleven to buy the essentials when on a punting trip?

https://www.todayonline.com/worl ... ognition-technology

This is actually my area of expertise so I’ll weigh in.

First, it isn’t facial recognition unless there is a reference sample from which to identify you. It is just facial caputre.

If you don’t opt in to provide your info, then they have your face, as do about 1,000 other cameras you walk by each day, but they can’t reference a sample with your specific data.

Hit rates in many progressive face capture technologies are currently at or above 95% in capture when properly installed and calibrated, this is no simple act either. There are vast number of variables that either enhance or degrade capture rates in current systems.

Many current technologies leverage what is called progressive assessment of faces. Basically, this means once an initial face is captured, it is cataloged and in the future if there is some level of abhorrent behavior by a person that capture is compared to the cataloged faces for a “match”. The matching is a variable set by the system operator that can provide a accuracy scale on the search criteria.

So you can say, show me all faces that are a 80% match or better and you get X number of results. Change it to 60% match and you get 5 times as many posssible matches. If you find a match from the system recommendations and you link it, the system now learns that face and get more accurate in the next search for that specific face. The more human verified matches a face has, the greater the likelihood that eventually the system can do live recognition searches for that face in the future.

Basically, unless you do something to become a reference sample, you will never know this tech is even in use. Currently in the US there are several retail companies that capture well over 1 million faces per day. They just don’t advertise their use or ability to the media like this. I have no doubt this was done by the technology vendor and 7-11 is not happy they did it.

Thanks Jake. But you would become a reference sample the moment you walk into a 7 eleven?

What I got from the article was that the system stores you in store behaviour including which racks you spend most time looking at into a database. The key to retrieve that info is your face. So staff knows your typical buying pattern the moment you are in the store and can offer you promotions.

Plus the tech seems Chinese, so next time you visit China the LE there would have a pretty good idea about your Thai trip.

looking at what you look like ... the answer must be most definitely for you  

I would be happy if this technology could ring an alarm bell every time I go to buy a Kit Kat when pissed at 2am in Bangkok.  

Obe

Your captured image is just that, data. Unless you pay by card, there is no connection to you.

Current analytics limitations make what you are suggesting a near impossibility just from processing power alone.

If a store wants to evaluate a specific product’s sales demographics, then your image could become an initial reference sample, but then it is about the product and not you the person.

The tracking of VIPs is again only if the individual is an opt in for the program. This is nothing new. Many casinos have been using face capture to identify two key groups in their properties for years.

Sharks. The big spenders that love extra TLC during their gambling sessions. They are captured at the door on an analytic capable unit, their entry is validate by a human, the vaild hit are then pushed to guest relations and a specific response protocol is initiated to enhance the shark’s experience at the property.

Cheats. Almost all high risk cheats are well known in any major casino. The property has high resolution reference samples of the highest risk cheats and the system is always on active search for these faces.

These are active reference samples that are individual based and have a rather small population size compared to all captures in a given day. This is where the processing is being used most proactively today in any industry.

Your trip to buy your girlfriend tampons will be very unlikely to result in your face ever becoming a reference sample.

One deterrent that isn’t called out and should be kept in mind is that many of these are in place to help solve future crime events. If I am doing capture in a location that is later the victim of a crime, I am easily able to go back and review all previous captures for high match comparatives. One thing that is found to be true in many crime incidents is that the perpetrators are often in the business prior to the actual crime being committed as normal customers.

Just my .02.

J

[ Last edited by  jake.houston at 21-3-2018 09:39 ]

Interesting timing... there was a great article on CB Insights yesterday about China Surveillance State and the AI startups that are part of it!

https://www.cbinsights.com/research/china-surveillance-ai/

China is pouring a TON of money into AI.