fashion 2017 hijab

thank you very muchfor coming to this talk. it's going to be aboutthree general topics. computer vision,surveillance and camouflage. if you have been to gatwick airport,you may have seen one of these. it a facial recognition system. designed to improvethe flow of traffic around airports. according to their website, the system is designed to capturepassenger’s facial features as they enter the airport.

and tracks them throughout the journey. it claims 10 times the capture rateof device tracking solutions, such as tracking the mac addressof your wi-fi. and the passenger of coursenever needs to switch on the technology, because your face it is always on. the company is interesting,but not very unique. they are part of a rapidly evolvingindustry of human analytics. on their websiteyou will see this statistic. they mention that 89% of peoplewould give biometric information,

as they travelthrough international borders. this is a statistic that you seeon a biometric industry website. however, this is a statisticthat you would not likely see. this is a view from a study by accenture. it's says 75% of respondentswould not even shop in a store that uses facial recognitionfor marketing purposes. the cameras look new and innovative,face detection and recognition isn't new. the history of face detectionbegins around 1969. when three japanese researcherspublished the paper

proving that a computer could be usedto detect a human face. it looked something like this. what you see is basically edge detectionbeing used to trace the outline of human silhouette. being able to match a unique silhouetteto the silhouette of a human was what ledto the first face being detected. following that breakthrough in 1969,in the early 70s researchers at stanford followed by another researcherat a japanese university published the first research that showedhow to do facial landmark tracking.

being able to detect the eyes,nose, mouth and shape of the head. this facial landmarking iswhat is used today by facebook or other companies to dohigh-performance facial recognition. you can see in this imagehow the algorithm in 1973 was able to segment the eye,the lips and the nose. if you compare this number from1973 of about 75% accuracy to the numbers in 2016. it is interesting to notethat computers are now at around 98.5% successin recognizing and classifying humans.

but meanwhile, humans are stuck at 97.53. in 2017, the number for humansis going to stay the same, the numbers for computerson the bottom will keep increasing. certainly, 2016 is an interesting yearand exciting for facial recognition. it really depends onwhere you're standing. it depends on whether you are someone using facial recognitionto collect or process information. or if you are being watched by it. toolkits have been recently released,both are called open face.

they are a little bit different. the one on the leftis a deep neural network that uses, i think it gets between 97 or 98% successin classifying faces and it's free. the one on the rightis used for behavior analysis. while the one on the leftis used to read the outside of you, to use you as an index of your identity. the one on the right readsthe inside of you, the emotional state. both could be deployedfor the cost of a raspberry pie kit, so around $50, to implement

a high performancefacial recognition system in 2016. this line actually comes from a marketing newsletter,from marketingland.com. they are very excitedabout facial recognition. in the marketing business, one of the problems isthat people like to pay in cash. i noticed here in berlin especially,a lot of things are cash business. you can't track that.facial recognition changes that. what marketers are doing,is using facial recognition

to make it possibleto track cash purchases. when you think aboutthe history of facial recognition, how it was originally founded by the department of defensefor military purposes. specifically to locateenemy combatants and criminals, the same technology is now usedto identify and track consumers who do not pay with a trackable paymentlike a credit card. that's not great. i wanted to provide a quick overviewof some capabilities in computer vision.

that have become very easy to deploy on anything as simpleas a raspberry pi surveillance system. the first oneis using gender and age detection, although you can seeit's not 100% accurate. it's open sourceand accessible to anyone. another one is called gaze detection. it looks at what you are looking at. if there is a photo of you, this algorithm can tellwhat you find interesting in that photo

and you can drive analyticsand you can understand a scene. building off gaze detection, you can do something called"finding important people in images". when you know who is looking at who, you can determine the social orderof a group of people in photos. this is an image from a companycalled jetpack. they are using public pixel of instagramto build tour guides for cities. what we see hereis they are looking at facial hair like hipster moustaches and lipstick.

they used it as visual informationto say if you are visiting a city. to guide you to a place where there is a high concentrationof hipsters with facial hair. or a place where a lot of girlsare hanging out with lipstick. when you post photosonline on social media, these posts are the superfoodof artificial intelligence. these are feeding the algorithms which are trained to thensell products back to you. or provide gated levels of accesswith facial recognition technologies.

this is one i came across on reddit. someone was bragging abouthow they used computer vision. they scan facebook profile photos of overweight people they store their idand target ads to them on facebook. this is another one. this company affectiva, which claims to havethe world's largest emotion database. with 40 billion emotion data points. nearly 4 million faces analyzed.

compare that to the nsa: this company is scanningabout 5,500 faces per day. whereas when you look at the reportby laura poitress and james risen from the snowden documents,nsa is scanning 55,000 faces a day. you see,the commercial surveillance companies are nearly on the same fieldas government agencies. as i said, you can now usethis facial expression analysis, to read the thoughts of somebody,to know what they are thinking. profile their emotions.

there is a thingcalled facial action units. one of them is called lip suck,or jaw drop, or blink for example. every facial movementcan be read and analyzed. so i wondered what would happenif i ran this on a video. on the senate of intelligence committee. i hope we can this as ayes of no answer. last summer,the nsa director was at a conference, he was questioned aboutthe nsa surveillance of americans. he replied and i quote here:

the story that we have millionsor hundreds of millions of dossiers on people is completely false. the reason iâ´m asking this question is having served on the committeenow for a dozen of years. i don't really knowwhat a dossiers is in this context. i wanted to see if you could give mea yes or no answer to the question: does the nsa collect any data at allon hundreds of millions of americans? no, sir. it does not?

not wittingly. there are caseswhere they could inadvertently. a lot of these algorithmsare about 90-98% accurate. but i'm convincedthat this one is about perfect. reading it from another anglegives you a different answer. it goes from joy to disgusted. i would like to bring attention comparingcommercial and governments' surveillance. they aren't that different. sometimes they are the same thing.

this highlight attacksfrom a company called pitpat has given a blanket licenseto the government. pipat is notablebecause is the facial recognition company that was sold to googleand that's the one the nsa is using to scan faces,that are intercepted from communications. i think a face is one of themost powerful tools for communication. yet there are very fewif any acceptable ways to protect it from dubious or invasive kindsof surveillance. six years ago,

i started a project called cv dazzleas my thesis at nyu. we made this sketch,to see if it is possible to make some kind of hairand make-up appearance that would block face detection. it came up looking like this. each looks is unique,each look works the same way to block the face from being detected. none of these faces in the imagesare apparent to computer algorithms. yet to a human,it's obvious that there is a person here.

that's the key to this project. it is to explore this very fine linebetween what can a computer see, what can a human seeand what is socially acceptable? i think there is a very fine linebetween all of those overlapping circles. for the right occasion of course. so the definition is, that cv dazzleis a camouflage from computer vision. it uses these bold patternsand hairstyles to break apart important facial featurestargeted by face detection algorithms. the dazzle part comes from camouflageintroduced in world war i

introduced on ships, called dazzle. which was interestinglyinspired by picassos cubism. the that it works is, if you werelaunching ammunition at the ship, it's difficult to tell sometimeswhether it is one ship or two, whether it is going left or right. the idea is to confuse the opponentby using this bold pattern, to take apart the gestalt of the object. so similarly with these algorithmsthey can reverse engineer to see what they are thinking about.

this is a genetic algorithm reverseengineering progress, of open cvs. first of all,here you can see in the heat map there are some parts in the facethat are more important than others. with this informationyou can design a camouflage pattern that would targetthose most parts of the image. these are the facesthat are hidden inside, some of the face detection algorithms. kind of a foolish looking ghostin the machine. that is the frontal faceand profile face.

i give you some tips,what works and what doesn't work. what worksis creating asymmetry on the face, using dark hair against light skinand the other way around. altering the contrastor darkness around the cheekbone. a very important area is the nose bridgeand the area between the eyes, kind of the center of the face. in 2014 the new york timescommissioned new looks. those were developed to be betterperforming adapted to newer algorithms and i show you how these are measured.

but first, i show you a ground truth, to show what a face looks likewhen it is detected and how easy it is to detect. the red areas show an area that isdetected by the face detection algorithm. if you are doing military camouflageand it is painted all over the face, it does not mean you blockfor the facial recognition. because it is aspecific application to those targetedhighly sensitive areas of the face. when we did same tests on the imagescommissioned for the new york times.

you can see,there were no positive detections. the different rectangles representeach of the five face detection profiles. so on this it is working well, but you can seethat one of the five detectors sees it. it's important to consider camouflagenot as something making you invisible. that is the way it is oftenwritten about in the headlines. but camouflage is the idea, that you are understandingthe threshold of detection. and then creating some way of appearance

that lowers you just one stepbelow that threshold of detection. and that is doneby an analysis with computer vision with a specific algorithmthat you are targeting. since 2010, which is a while ago now,some interesting things have happened. there was an article, somebody saw a young woman with anti-computer vision,anti-surveillance make-up first sighting in the wild. anti-surveillance feminist poetry,hair and make-up party.

the band big dataused that for their promo. there had been art and science festivalsthat have done it and everyone interprets this differently. just using the tips on the website are able to constructunique camouflage designs that actually block millions of dollarsof face detection. and sometimes it is justbeyond my control what happens. this appeared in a tv showcalled elementary. in one of the scenethis criminal type character,

was trying to evade the face detection,so they created this look. which is quite colorfulbut this does function, it works. in other popular cultural referencesthe tv show "minority report", not the movie,imagined that in the future in 2065. it is kind of a throwback,where it was easy to block, all you had to do was using make-up. so this becomes something like a tattoo that maybe you regret when it's 2060 because it only functioned40 or 50 years ago.

some of the examples i have shown are more suitablefor going out at night, to a club. but it's worth pointing outthat it can look good. i wasn't involved with this, it is an example of a hair stylethat happens to work well. with the right hairstylistand make-up artist you can go with it. i want to shift now to my recent project. starting in 2012,it is about drone surveillance, in particularly military drones.

the capability ofthermal imaging systems like this camera on the bottom of the drone,with multi spectral imaging capabilities. one of the capabilitiesis to see in the dark or the thermal radiationthat is emitted from your body. it is hard to think of yourselfas always visible, even if you are hiding in the jungle,you can still be detected. that poses an ultra-thread to privacyand it's quite powerful technology. i wonder if this is really unstoppable. researched waysthat a thermal imaging could be blocked.

and in 2013 releaseda series of garments called stealth wear. these are garmentswith a metal plated fabric. and the metal is typically silver,so it is highly flexible, and you can seethat the metal plated fabrics reduce the thermal signatureor black out that part of the body. here is an exampleof somebody wearing it. you can see, there are 4 peoplebut there are actually 5. the projector makes it hard to see. once the individual starts moving,it is impaired.

but in the still imagesomeone is blending in perfectly to the background in the winter. that is more difficult, because you havea temperature differential. i'm going to play it,then it's pretty obvious. the intent is notto provide a full military solution but to illustratethe possibility for fashion that makes it more inappropriatefor the mass-surveillance. there is a hoody,there is a burka and a hijab.

these items are based on islamic dress. the idea with this project iswhereas religious dress is traditionally in some ways providing a separationbetween man and god, this collection reimaginesreligious dress in the context of mass surveillance as providing a differencebetween man and drone. some of the other interesting things,slightly concerning, were these emailsand some articles that came out. this project goes againstnational security interest.

which makes me as an artist quite aware of the possibilitiesthat go wrong. i got an e-mail in 2013 for a request, for publicationin a classified intelligence document which is strange. i will never see that,do you really need my permission? most newspapers don't even ask for it. the washington postput on a requested for comment on the officeof the director of the nsa desk.

i tried to create work that doesn'tput me at risk. the area of national securityis sometimes black and white and it is not an open area to art. but that is why i am doingthe projects that i do. the other thingis a tweet from the pentagon. just making note that they are aware of the ideaof hiding from military drones. there are grey areaswhere artists can operate. being able to not make projects

that are threateningfor any people in particular. but gives us room to think aboutthe future that is coming up. as a response to theanxiety provoking emails i received. the next project i createdwas the privacy gift shops, which is a way to makethe threatening idea of national security al little bit more friendly. because gift shops are always friendly,open, accessible, kind of safe places. this is a shop for counter surveillance,art and privacies accessories. i sell them directly.

the anti-drone hoodydid quite well and sold out. it is expensiveand i am a terrible businessman, i can tell you,you can do that for less than 1 dollar. the way the thermal radiation is blocked by fabricthat is reflective or insulating. so the most obvious examplewould be a space blanket. you can get a pack for your whole familyfor less than 10 dollars. while that has the same functionto block thermal radiation, the main differenceis the psychology of camouflage.

that is a space of my blanketthat is how this works. to sum things up a little: there are functional solutionsto hiding and blocking surveillance. and there are more artisticand psychological responses, more thought out kinds of camouflage. when we think of camouflage,it is vietnam era and 81 pattern. brown and greens and blacks. but camouflage didn't existthe way it does before the 20th century. in the early 20th century it was a termfor criminals hiding from the police.

i think you can draw similaritiesbetween the ways camouflage was thought ofin the early 20th century the way privacy was thought ofearly in this century. some other quotes from a book aboutcamouflage and its impact on australia. t. roosevelt considered camouflageas a form of cowardice, a mere defense of strategy. but as world war i and ii unfolded, he realized the significance and power,the strategic advantage. and camouflage became a signof humanities increasing intelligence.

privacy can be thought of the same waynot as a defensive response, but privacy is something that showshumanities increasing intelligence. in closing, we thinkthese solutions are eccentric, but if you look back in 1918in new york city, everybody was wearing a cap,there is one person that is not. if you look there now,nobody is wearing the cap. that is possible for something normalto become abnormal and the other way round in the future. thank you.

(applause) thank you adam.