Live data from Hacker News

A crashed advertisement reveals logs of a facial recognition system

twitter.com

21–30 of 550 posts

Re: A crashed advertisement reveals logs of a facial recognition system

#21
post #19

In Japan at least, before automated facial recognition, cashiers recorded buyer demographics by hand. I would think other places do it too. Edit: Here is what the buttons look like. Gender and age. https://image.slidesharecdn.com/hvc-c-android-prototype20141...

Could you be more specific? What could they possibly [Edit:] have recorded in a second other than male/female, Japanese/foreigner, or child/adult?

Facial recognition is a unique identifier but cashiers have access to almost nothing they can record... [Edit] What was it?

Edit: clarified that I am asking about the history here, what information was manually collected by cashiers as parent stated

Re: A crashed advertisement reveals logs of a facial recognition system

#23
post #10

Uh, I got sidetracked and brain hammered by the devolving discussion on that Twitter thread, thus couldn't find the context for this pizza shop kiosk - Is it a customer service portal that attempts to identify the person in front of it to try and match up with an order, or a plain advertising display that is trying to capture the demographics of the people who happen to stop in front of it and look at it?

It's an ad. May be targeted advertising, or just simple engagement analysis. http://www.dinside.no/okonomi/reklameskilt-ser-hvem-du-er/67... Edit: there is more information in the article. Not going to read it, though.

To summarize, it's an experimental project, there is so far only one such screen at the train station. If someone stands within 5 meters of the screen it will try to classify their age and gender and show a targeted ad, and record how long they looked at it. The raw images are not saved.

The screen uses a software called Kairos to analyze faces. It can estimate age, gender, and whether you are "white, black, hispanic, asian, or other".

According to the marketing manager at Peppe's Pizza, he thought there would a label on the screen saying what's going on, but in fact there is just a small sticker on the back of it, which is quite hard to see.

The company making the screen, ProtoTV, says that people should be okay with this because ads on the internet are even more targeted. A government representative says that the system might violate laws about surveillance cameras.

Re: A crashed advertisement reveals logs of a facial recognition system

#24
post #19

In Japan at least, before automated facial recognition, cashiers recorded buyer demographics by hand. I would think other places do it too. Edit: Here is what the buttons look like. Gender and age. https://image.slidesharecdn.com/hvc-c-android-prototype20141...

Could you be more specific? What could they possibly [Edit:] have recorded in a second other than male/female, Japanese/foreigner, or child/adult? Facial recognition is a unique identifier but cashiers have access to almost nothing they can record... [Edit] What was it? Edit: clarified that I am asking about the history here, what information was manually collected by cashiers as parent stated

Demographics and sales data is a primary example of big data.

The tweeted advertisement system also looks like it's only recording demographics. Not individual personal IDs.

Re: A crashed advertisement reveals logs of a facial recognition system

#25
post #20
post #12

Though not at the level depicted in the movie, I am nonetheless reminded of Minority Report.

For those that haven't seen it, here's the scene where all the digital signage knows who John Anderton is: https://youtu.be/7bXJ_obaiYQ The movie was released in 2002, so fairly prescient.

I thought of the film as well. I think we should remember that for all the cool cars in the setting, it was meant to be dystopian.

Re: A crashed advertisement reveals logs of a facial recognition system

#26
post #19

In Japan at least, before automated facial recognition, cashiers recorded buyer demographics by hand. I would think other places do it too. Edit: Here is what the buttons look like. Gender and age. https://image.slidesharecdn.com/hvc-c-android-prototype20141...

Could you be more specific? What could they possibly [Edit:] have recorded in a second other than male/female, Japanese/foreigner, or child/adult? Facial recognition is a unique identifier but cashiers have access to almost nothing they can record... [Edit] What was it? Edit: clarified that I am asking about the history here, what information was manually collected by cashiers as parent stated

You could distinguish people who are married or are parents (with false negatives) by recording people who are at the till with their spouse or children. (Send out demographic-research cards once, to a few of the same stores you've collected this info from, to derive a normalization factor that will make such collected observations useful from then on.)

You could make a note of a person's seeming affect—positive/negative/neutral emotion.

Re: A crashed advertisement reveals logs of a facial recognition system

#27
post #24

Earlier quoted context omitted.

Could you be more specific? What could they possibly [Edit:] have recorded in a second other than male/female, Japanese/foreigner, or child/adult? Facial recognition is a unique identifier but cashiers have access to almost nothing they can record... [Edit] What was it? Edit: clarified that I am asking about the history here, what information was manually collected by cashiers as parent stated

Demographics and sales data is a primary example of big data. The tweeted advertisement system also looks like it's only recording demographics. Not individual personal IDs.

Sorry, I wanted you to expand on this: "cashiers recorded buyer demographics by hand".

Give me an example of what they used to record by hand. All I can think of is "male, adult". I am specifically interested in what else you say they used to record.

I wasn't asking about the present status quo, only your historical statement about cashiers recording by hand.

Re: A crashed advertisement reveals logs of a facial recognition system

#28
post #20
post #12

Though not at the level depicted in the movie, I am nonetheless reminded of Minority Report.

For those that haven't seen it, here's the scene where all the digital signage knows who John Anderton is: https://youtu.be/7bXJ_obaiYQ The movie was released in 2002, so fairly prescient.

"Your video will play after this ad" oh the irony

Re: A crashed advertisement reveals logs of a facial recognition system

#29
post #24

Earlier quoted context omitted.

Demographics and sales data is a primary example of big data. The tweeted advertisement system also looks like it's only recording demographics. Not individual personal IDs.

Sorry, I wanted you to expand on this: "cashiers recorded buyer demographics by hand". Give me an example of what they used to record by hand. All I can think of is "male, adult". I am specifically interested in what else you say they used to record. I wasn't asking about the present status quo, only your historical statement about cashiers recording by hand.

Oh, OK. I edited in a link to the buttons the cashiers used, in the original post.

Re: A crashed advertisement reveals logs of a facial recognition system

#30
post #26

Earlier quoted context omitted.

Could you be more specific? What could they possibly [Edit:] have recorded in a second other than male/female, Japanese/foreigner, or child/adult? Facial recognition is a unique identifier but cashiers have access to almost nothing they can record... [Edit] What was it? Edit: clarified that I am asking about the history here, what information was manually collected by cashiers as parent stated

You could distinguish people who are married or are parents (with false negatives) by recording people who are at the till with their spouse or children. (Send out demographic-research cards once, to a few of the same stores you've collected this info from, to derive a normalization factor that will make such collected observations useful from then on.) You could make a note of a person's seeming affect—positive/nega…

I'm asking historically what was actually done, not what could be done. (For, "what could be done" you could ask if the cashier had seen this person before? Are they a regular shopper here, as far as the cashier notices?) I am asking what information cashiers in Japan actually in fact recorded by hand. What did these cards look like for each customer. Etc.
Post reply on HN