I ran this same image through our face detector, with results I could easily predict: http://demo.pittpatt.com/detection_demo/view.php?id=JSCV5AH9... We don't find any of them (blue and yellow are low confidence faces). We do "well" by our definition, terrible by the writer's.
So, problem #1: using hand-drawn faces (these are fairly stylized) is a really bad way to test a face detector. No one "in the real world" wants to detect or recognize hand-drawn faces, and so no one trains with hand-drawn faces. We, specifically, err on the side of choosing to exclude stylized faces from the faces category (though most frequently these would be "don't-cares").
Problem #2: Using a single image is actually a bit misleading, because if you really want to "test" methods for thwarting detection, you need to use a video. Those slight variations in pose and lighting are going to make it much easier to pick up the face and filter out the misses.
Lastly, if you wanted to "defeat" a dystopian mass surveillance system, you don't want to prevent detection, but recognition. It's far easier. And to prevent good recognition over a huge dataset (ie, the population of the world), you just need to "remove" information from your face. Wear big ol' sunglasses and a hat. Far more effective, far less conspicuous.
(edit to add: if you click 'thesis' at the top, and read the top few entries, he shows alot more about how viola-jones and the haar wavelets are used to find face regions. It's an interesting visualization. It also does a great job explaining why the makeup trick works)
The headline on this HN submission is a bit ambiguous: are the face-detection systems using the Viola-Jones method, or the the face-hiding technique a Viola-Jones method?
Honestly, I doubt it. They'll remember the makeup probably, and likely little else. Unless your hair is also noteworthy, I doubt most would remember it.
I remember the story of the naked bank robber, who got away with it because nobody could describe his face, as they were too distracted by his nudity.
I disagree, anybody who is doing face recognition for a serious purpouse is probably using lots of surveillance cameras. In the setting where you can make someone take off makeup it's probably easier to just ask them for their credentials.
Agreed, however I haven't seen anyone mention the more obvious (or maybe not so obvious) point. That sort of disguise might evade the algorithms, but it would be glaringly obvious to any person. It's rare to have surveillance 100% managed by cameras and computers alone, there is usually some amount of physical guard presence. Even a semi-catatonic guard would likely investigate a person made up like that moving through a crowd. Especially if they were male.
Fantastic! I just always knew that in our distopian cyberpunk future, we mercenary code gargoyles would all be covered in some kind of digital war-paint-camouflage as we wage secret mercenary wars on behalf of covert multinational-mega-city-corps.
So, problem #1: using hand-drawn faces (these are fairly stylized) is a really bad way to test a face detector. No one "in the real world" wants to detect or recognize hand-drawn faces, and so no one trains with hand-drawn faces. We, specifically, err on the side of choosing to exclude stylized faces from the faces category (though most frequently these would be "don't-cares").
Problem #2: Using a single image is actually a bit misleading, because if you really want to "test" methods for thwarting detection, you need to use a video. Those slight variations in pose and lighting are going to make it much easier to pick up the face and filter out the misses.
Lastly, if you wanted to "defeat" a dystopian mass surveillance system, you don't want to prevent detection, but recognition. It's far easier. And to prevent good recognition over a huge dataset (ie, the population of the world), you just need to "remove" information from your face. Wear big ol' sunglasses and a hat. Far more effective, far less conspicuous.
(edit to add: if you click 'thesis' at the top, and read the top few entries, he shows alot more about how viola-jones and the haar wavelets are used to find face regions. It's an interesting visualization. It also does a great job explaining why the makeup trick works)