Live data from Hacker News

Clothing designed to confuse facial recognition software

capable.design

11–20 of 72 posts

Re: Clothing designed to confuse facial recognition software

#13
The current HN headline [0]makes it clear what this is, but the actual website makes it really difficult to understand what their value proposition is, other than “designed in Italy made from Egyptian yarn”

[0] “Clothing designed to confuse facial recognition software”

Re: Clothing designed to confuse facial recognition software

#15

I don't know anything about facial recognition software so I'm probably underestimating the difficulty of alerting the system and it's operators to the interesting possibility of giraffes and zebras wandering around the urban environment.

Facial recognition tech is basically two parts: detecting a face in an image (“face detection”, often this is done by calling out to OpenCV’s face detection in Python or C++), and then extracting information from that face image and searching a database with it (“facial recognition”, sometimes done with algorithmic measurement of facial features but increasingly done by neural networks).

This clothing looks like it might be designed to confuse the face detection step.

Re: Clothing designed to confuse facial recognition software

#16
post #5

I love the initiative, but are there any... less gaudy designs?

I think they have to be gaudy to fool cameras.

I doubt this is true. I've seen adversarial version of images which are very similar to the original image.

E.g. an image of a cat that has had some pixels flipped so it's classified as a dog instead.

Re: Clothing designed to confuse facial recognition software

#17
post #14

This seems to prevent people from being detected by a generic object detector, but it doesn’t seem obvious that it will actually confuse a face detector… it might however confuse an autonomous vehicle, wouldn’t wear this in the street

I’m not sure where you get that from. These are specifically designed to confuse face detection software.

Edit: Maybe I see the cause of confusion. The videos show “person” detection, but the way these systems distinguish people from other objects is by faces. As far as I know cars don’t do that, they just detect objects and don’t care about faces, so it shouldn’t be an issue.

Re: Clothing designed to confuse facial recognition software

#18

The current HN headline [0]makes it clear what this is, but the actual website makes it really difficult to understand what their value proposition is, other than “designed in Italy made from Egyptian yarn” [0] “Clothing designed to confuse facial recognition software”

The web site copy is terrible! Every sentence vaguely hints at what the product is for without outright saying it.

> Cap_able offers a high-tech product that opens the debate on issues of our present that will shape our future. Cap_able wants to have an impact on society, creating awareness on contemporary issues through highly innovative design products from a technological and ethical point of view. [and even more of this]

Ok, ok, but what… is… it…?

Re: Clothing designed to confuse facial recognition software

#19
post #14

This seems to prevent people from being detected by a generic object detector, but it doesn’t seem obvious that it will actually confuse a face detector… it might however confuse an autonomous vehicle, wouldn’t wear this in the street

Yes, the only mention of software is Yolo (wrong capitalization and no mention of version), and I can’t find a link to a study or anything like that. Not convinced it can meaningfully confuse a classifier specifically trained to recognize faces, not giraffes or zebras. Sounds more like marketing BS.

Re: Clothing designed to confuse facial recognition software

#20

Earlier quoted context omitted.

I think they have to be gaudy to fool cameras.

I doubt this is true. I've seen adversarial version of images which are very similar to the original image. E.g. an image of a cat that has had some pixels flipped so it's classified as a dog instead.

Something like flipping a pixel would work if you manipulate the actual image (raw image) being fed to the software.

Here it is captured via a camera and then fed to the software and I doubt pixel level details are captured by a camera

Post reply on HN