27
votes
Can data die? Why one of the internet's oldest images lives on without its subject's consent
Link information
This data is scraped automatically and may be incorrect.
- Title
- Can Data Die? Tracking the Lenna Image
- Authors
- The Pudding
- Word count
- 504 words
This post showing up here and now is quite the coincidence for me. I just finished reading Emily Chang's Brotopia: Breaking Up the Boys' Club of Silicon Valley, and the book opens with the history of the Lena image and uses it as a stand-in for the tech industry's treatment of women:
This was the first I had ever heard of the image.
Also, this is nothing more than a petty complaint, but I find it frustrating that the Losing Lena documentary, which is focused on the topic of data-removal, is hosted only on the, presumably, very data-hungry Facebook Watch platform.
I became familiar with the image because it was used as an example in early computer graphics papers. At the time, if I thought about it at all, I probably assumed it was a stock photo, like the photos of good-looking strangers in Kodak advertisements. I didn't find out the history, that it was originally from Playboy and cropped, until a lot later. (Racy!) Most likely, new papers used it because older papers did and good quality scanners weren't common yet.
I can give you a couple of other examples of "famous" computer graphics datasets. There is a teapot in the Computer History Museum in Mountain View. They used a teapot because it was a relatively complex shape (at the time), but it's just a teapot. As 3D computer graphics got more capable and the teapot was no longer a challenge, researchers switched to the Stanford Bunny.
Nowadays, ImageNet is a famous dataset of over 14 million images that's used by machine learning researchers. It seems that researchers are only marginally more respectful of copyrights?
It seems like there are two issues here. The first is copyright infringement, which is rampant on the Internet. I would bet that many people who are concerned about Lenna have also shared memes with people in them without a thought about whether those people consented to be a meme. Stopping meme sharing, or even getting consensus that it is a problem, would be pretty difficult. (For one thing, you have all the people who say DRM is evil. But without some sort of computer assistance, making a big dent in infringement seems hopeless.)
The second is a vague complaint about it being objectifying, which I'm not sure how to think about.
For example, it's common for a camera review site to demonstrate a camera using photos of a good-looking model to demonstrate camera quality. (Presumably with their consent.) Occasionally I've wondered about the strangers who posed for those photos. Are they a friend or co-worker of the camera reviewer, or did they hire someone just for the shoot? Is wondering about them being less objectifying than if I was just comparing zoomed-in image details to see which camera is better?
It seems kind of wild that there are articles and even a movie about this particular image, versus the zillions of other images out there on the Internet. Fame leads to more fame, I guess.
I’ve definitely seen that teapot before — multiple times, I think — but I can’t put my finger on where! My instinct is videogames? It’s definitely pinging for me though, and I had no idea it was significant until now.
And yeah, as for the issues you brought up, I don’t think we’ll “solve” copyright infringement online (assuming it’s even a problem in the first place). Sharing, modifying, and reappropriating content are so fundamental to internet culture and structure that we’d essentially have to fundamentally rework things from the ground up and reacculturate users to get anywhere on that front. It feels like this might be slowly changing recently, but for nearly all of the time I’ve been online, one of the easiest ways to get something spread around online is to tell people they can’t or shouldn’t share it. I still remember people copying that DVD encryption key to every site under the sun as an act of social justice, for example.
As for your second point about objectification, I don’t want to speak too definitively on it, as this one’s far out of my lane, speaking as both a) a guy and b) someone with no experience in the tech industry.
With a grain of salt on mandatory offer, my read of the Lena image (which is based solely of the linked article and it’s mention in Brotopia) is that the image itself isn’t the issue so much as it is a measurement of the issue. When one of the most prominent and well-known woman in the industry isn’t an actual woman but an image of one, that says something about valuing women primarily at an aesthetic and functional level (i.e. how Lena’s image helped develop image processing algorithms). Furthermore, when Lena is in spaces where she is the only woman or even one of a small handful, it acts as a measurement of a lack of women in the field. The problem isn’t Lena herself but what her status says about the status of women in the industry at large.
These two issues can both exist independent of intention, and they can reinforce each other. A culture which objectifies women can fail to prioritize their hiring or even drive them out when they are hired, and a culture which lacks women can lack the perspectives and social experiences that would reduce their objectification. This is one of the main complaints within Brotopia, which goes into the issue well beyond the Lena image itself (which is only a minor jumping off point for the beginning of the book).
That said, some people do take issue with Lena herself, and I don’t think that’s an invalid read either. I don’t think women in tech should have to be continually reminded, in everyday professional settings, of their own objectification in the form of pornography. Once you know the backstory of the photo, it’s hard to unknow, and if Lena really is omnipresent, then she becomes a continual reminder. If her image is what’s giving a Hooters-type overtone to what should otherwise be a family-type restaurant, then I can see why some people would be uncomfortable and would want to reevaluate Lena’s role, prominence, and necessity moving forward.
But again, take all of this with a big grain of salt. You worked in the industry yourself and have direct experience with the image in question, while I’m just looking at this through a window, peering in curiously from the outside.
I see that the teapot has a Wikipedia page, as does the Lena image.
I’m not a real computer graphics expert. I took a course in college, skim academic papers sometimes, and went to SIGGRAPH once back in the dotcom era. It’s enough to recognize some of the in-jokes.
I don’t believe there is any serious reason to use an old 512x512 test image anymore. In newer papers that are about portrait image processing, there is typically a grid of photos of people with a variety of hair styles and skin tones, showing each original photo and what it looks like after processing with various effects.
It doesn’t seem to be the custom to say in the paper where the images came from or who those people are, but I wouldn’t be surprised if sometimes they are pictures of the researchers themselves? They look like they might be grad students or professors.
The image appears in this article too. I wonder if they asked for consent, or assumed it’s fair use?
… how else would you describe what an image on a computer is?
Initially I thought they meant it was a bunch of coloured
divelements or similar, one per pixel, but looking at the page it's actually a singlecanvaselement - which ironically gave me "Save image as..." as the first option when I right clicked to inspect it.I see the intent there: it's harder to link from elsewhere, pick up by crawlers, etc. but if obfuscation was the goal I think they could've gone a step further, and if preventing reuse was the goal then maybe a watermark would've worked better.
More broadly I get the impression that context matters here, and that talking about the choice of image and its implications is both legally and ethically reasonable in a way that using it as a placeholder isn't. Given that you pretty much can't remove something from the internet, much as we might wish otherwise, the slight obfuscation seems more of a symbolic gesture than anything else.
Yeah that's a rather odd thing to say - any raster image is a collection of colored pixels. Maybe they mean that it's pixelated?
Ah, thanks! Missed that.
Yeah it looked cool and it was nice that they could sync the graph with the text, but it wasn't comfortable to read at all.
Good lord I hate the design of that site. Thank goodness for reader view.