Hacker News new | past | comments | ask | show | jobs | submit login
LenPEG (2017) (dangermouse.net)
78 points by tosh on Jan 9, 2021 | hide | past | favorite | 44 comments



LenPEGs average compression ratio has sharply fallen since journals started discouraging the use of Lenna as a test image

The field is ripe for disruption by finding the new optimal image to use as the special case


For anyone who doesn't know... Although the Lenna reference image has historical significance within some fields (and that's part of the joke of the article), you probably don't want to use it in your work today:

> In 2018, the Nature Research family of journals announced that they would no longer consider articles using the Lenna image.[19] In the same year SPIE, the publishers of Optical Engineering, also announced that they would discourage the use of the Lenna image, and would no longer consider new submissions containing the image "without convincing scientific justification for its use".[20]

https://en.wikipedia.org/wiki/Lenna#Criticism


I remembered when I was interning about 10 years ago at Lockheed Martin doing computer vision work, we were using the Lenna image as a testbed image inside Matlab. I remember when I was installing the repo, my manager told me to use the Lenna image, and I asked him what it was, and he said "don't google it on a work computer!".

He explained to me the origins, and actually gave me the image file from a flash drive, just to minimize the likelihood of me getting in trouble. While obviously the normal Lenna image itself is fairly innocuous, he was afraid that porn might come up if I were to look up anything in Google images, which probably would not look great on an intern's computer.

In hindsight, maybe it's best to not use images that could risk getting anyone in trouble, not to mention that it probably breeds a "boy's club" atmosphere.


That's pretty sad, I remember seeing the Lenna image 20ish years ago when I first looked into computer vision, I was a kid back then and I dreamed of one day publishing a paper with this image in it (that's how ubiquitous it was, to me it was the incarnation of CV). I used it often on my website just a few years ago to demonstrate CV projects. I don't find it sexist, I'll keep using it.


I'd like to ask you to reconsider, and here's why...

I imagine that the image was originally used with a sense of whimsy, with no harm intended, and then it was adopted into the culture of some fields, and given new meaning there within those fields, however we now know (at least when it's pointed out) that it does send some very unwelcoming messages, which affects both the fields and broader society.

I think one of the unwelcoming messages that would be sent by continuing to use the image is: knowing that it's associated with a long history of serious exclusion and unfairness, were we to use the image anyway, that would say we care more about the field's old in-joke, than we care about injustices and about being welcoming to everyone interested in the field.

This seems like one of those situations in which we don't realize something is a problem, but once we realize it (maybe someone has to mention it, and we look into it), then we do what we can to correct it. In the most limited interpretation of the problem, the correction might be as simple as using a different image in our examples, and it's perhaps the very least one can do.


I've considered it carefully. I didn't expand on it because I know my opinion is usually not welcome (eh) here and also there is no way to come out looking good when arguing with people who position themselves as fighting against prejudice even if you care about social issues and it's exhausting.

As you say, it's the most limited interpretation of the problem. It's to the level of the master-branch or black-list/white-list controversies. I believe sexists are unwelcoming (to say the least), and sexism breeds a toxic work environment. I'm sure this image has been used by sexists, as has every resource in every field up to this day. I do not believe this image is sexist or unwelcoming nor that changing it will do anything to improve the work environment for women in tech.

I think these requests are overall harmful: They're a rhetorical mirage that does nothing to help the people they claim to stand for while disenfranchising normal people in the process by reframing them as bigots. They are part of a broader pattern of ideologically motivated battlegrounds in (SV) tech and serve only to classify people as allies or enemies. You yourself frame it as dichotomy between people who care about injustices and people who care more about preserving an old joke (clearly assholes if they put a joke over the inclusion of their colleagues). I'd ask you to consider that people who share your beliefs do not have a monopoly on kindness and mercy and the human heart, and to consider furthermore, given the polarization of the American political discourse, whether this is the right way to build a welcoming environment.

I say all this from personal experience, as a minority who's unlucky enough to have the most privileged people on earth fight for him (but not really) one well-meaning frivolity at a time. I say unlucky because in the process every space for us online has been hijacked and paradoxically I have less of a voice than before and meanwhile everything is still pretty shit to be honest.


I would ask on what basis you think this image is not sexist or unwelcoming? You are allowed to think that of course, I'm just curious.

I have a small sample size but I have been in a conversation with my wife and some of her female friends (who all had whom have CS degrees), and they all agreed they dislike the use of this image.


Discussions like this feel very American to me. From a European exaggerated POV, everyone over there wants to either be a victim or fight against discrimination towards a group they are not part of, which always feels like people just seeking attention or making yourself feel good like a white savoir, without solving an actual problem.

At least you actually asked women about their opinion on this image. Fwiw, I just sent the image to my family group and only got two replies so far which were both just a confused "huh why?". To be fair, both are not in CS or even engineering.

Again I think this is due to cultural differences. The image doesn't contain nudity or suggest anything sexual to us over here. It's a woman with an attractive face. A prime example for the difference in perception is how (gun) violence and nudity are perceived in Europe vs the US in movies or games, it's pretty much reversed.


I am in Scotland.

What did you ask. Was it something like "Do you think it might put girls off computing to know for years the main picture used for image processing was a cropped picture from Playboy" (I don't mind you asking a different question, but I'm curious what you asked).

Yes, the image "as is" is fine, but people in the area know where it comes from, and what it is used for.


I asked whether it should be controversial or sexist to use this cropped picture from Playboy as a sample image in computer science. No additional background info or context.

In general the Playboy is not considered "dirty porn" here in Germany which at least in part can be credited to the student movement of the late 60s which among many things addressed sexual self determination. So by some, a woman deciding to be featured in the Playboy or even porn was not seen as degrading but rather the opposite.

What might further surprise people from other parts of this planet is that for example at supermarkets and stores, the section featuring magazines has your typical tech, gaming, glamour magazines right next to ones with boobs on the cover for everyone to see. You see this from childhood on, it is not something bad and horrible and you're not taught that these women are desperate and couldn't find any other way to make money to survive.


I am quite aware that women can be empowered and naked -- that doesn't mean it fits well with tech.

I'm glad women in your life are happy with this, really. I really don't think this is a problem with naked women being shameful, but I really do think it's not appropriate for children. Do teenage girls buy playboy in Germany? I imagine usually not.

Honestly, we might have to agree to disagree, but I really do think using Playboy as a standard image gives the impression this is "for boys", the same way Playboy itself is generally seen as "for boys".

If this was just about nudity being fine, I am not seeing many naked men in tech.


> Do teenage girls buy playboy in Germany? I imagine usually not.

Not that I'm aware of, neither do I know in general which and how many women consume similar media in general. My impression, and I'm not feeling very confident in suggesting this is correct or representative, is that women in general rather like to talk about sexuality related topics with their friends, or for example check out guys together on Facebook etc and not so much look at nudity in a magazine.

> Honestly, we might have to agree to disagree

I'm totally fine with that. :-) I'm not on a mission here, just trying to explain why I feel this is rather an overreaction, and how my cultural background might play a role here... Yes it came from playboy, but the image on its own barely hints at that. At least I never made the connection and only learned about it from the HN comments a few years ago, I think from an IOCCC entry that used that image in some way.

> If this was just about nudity being fine, I am not seeing many naked men in tech.

I'd be absolutely fine with someone using a cropped face from a Playgirl model as a counter weight. :-)


People can be sexist or unwelcoming. An image is just an image, it's an inanimate object, it doesn't have opinions of its own. It's the meaning one attaches to it that can be sexist or not - so it's in the eye of the observer. To me, women and men have the same dignity, conferred by being a human being, so I don't see a picture of a woman as sexist (in the same way that I don't see a picture of a man as sexist). Even nude art images are not sexist. Of course there have been people seeing, even using, images in a sexist way. But this should lead us to condemn those behaviours and those perpetrators. Otherwise it would be akin to saying that kitchen knives are evil because sometimes they have been used to kill people. No, kitchen knives don't have any moral connotation, it's behaviours and people that can be evil (and usually even then, the same person is not pure evil all the time).

EDIT: having said this, if an image is making some people uncomfortable, we can use a different one. But in my view this seems just avoidance, a quick way out, which is going to make the original problem worse in the long term, because sexist people will remain sexist, only less discernible, in "stealth mode".


I agree a single naked image of a woman is not sexist, but why is it that almost every semi-naked image or person I see in tech is female? They are used in advertisements and as "booth babes". If the image wasn't sexist, then there should have been a 50/50 chance of the standard image we use for images being a clipping of a naked man -- but I believe that would never have become a standard image used in CS. I can't prove that, but I strongly believe it.


Regarding tech, I guess it's because it's still a field where (unfortunately) there are more men than women, and since men are attracted to women, advertisement agencies use this fact to capture the attention of the audience. It's these ad people that are sexist or at least using sexual attraction as a leverage. The day a field will have more women than men, they could switch to using semi-naked male imagery (even though there could be different dynamics in play that dictate the rules). The effect is unfortunate of course, because a male dominated field becomes less compelling for women and thus creating a negative spiral.


> They are used in advertisements and as "booth babes"

Hasn't this largely gone away in recent years?

It's still the norm in certain industries (e.g. online casinos), but my impression is that the average tech company doesn't want to be seen as the kind of organisation that hires 'booth babes'.


Very well said. I have the impression that some of these fights against injustice are turning into fights against a set of labels and are effectively just sweeping the real problems under the rug. It's very well possible to be a racist even when totally avoiding the use of master-slave terminology in software, as it's possible to be inclusive and non-racist when using it. Having a binary view on concepts or people is never good in my opinion.


Sounds like you have a thoughtful and nuanced rationale. I'll have to think about it.

For now, I'm wondering what the effect would be, if we were all, e.g., still using the Lenna image in things we publish. Would that be discouraging to people, regardless of what nuanced rationale we could in theory give for why we continued to use it?

Also, when inquisitive children read ahead and learn on their own, and come across these articles, would they, e.g., pick up on and be influenced by outmoded, restrictive gender roles?


Your argument makes sense, but I'm grasping for help understandingone thing.

> it's associated with a long history of serious exclusion and unfairness

I might be thick now, but could you explain the association, and how using it means that we condone this history?


In my (short opinion), that tech is a sexist boys club, and that for many people the use of porn backs up this idea -- else why not use anything other than porn? Or also a clipping of a picture of a naked man?


I understand that it is a pretty-ok image for playing with image processing stuff, but do you not think that using literal porn might breed a bad atmosphere in a professional environment?

I'm definitely guilty of using the image, I still occasionally use it on my home computer if I'm playing with image processing stuff, but I purposefully do not use it for anything that any other human is likely to see, since I don't want to be the reason to make someone feel unwelcome in the tech space.


Even if you overlook the sexism angle, it's always been legally dubious since Playboy owns the copyright and never gave permission for it to be used by researchers.

Plus the canonical 1970s-era digitization is low resolution and lacking in detail compared to modern digital photography. Take all three factors together and it's clearly time to move on.


Somehow I doubt Playboy cares about a cropped image that's been used everywhere for the past 50 years, as evidenced by their lack of action for 50 years. If the image is too low resolution to showcase your work, you would not use it in the first place. The resolution was good enough to showcase a whole range of state of the art work I was reproducing 5 years ago though.


I doubt Playboy cares either, but given the choice between "Playboy could sue us but probably won't" and "just stop using that old porn photo already" I can see why journals legal counsel would push for the latter.


'The editorial in the January 1992 issue of Optical Engineering (v. 31 no. 1) details how Playboy has finally caught on to the fact that their copyright on Lena Sjooblom's photo is being widely infringed. However Wired mentioned that: "Although Playboy is notorious for cracking down on illegal uses of its images, it has decided to overlook the widespread distribution of this particular centerfold".'

'"We decided we should exploit this," a Playboy rep said told Wired in 1997, "because it is a phenomenon."'


> I don't find it sexist, I'll keep using it.

How do women in your industry feel about it?


even if you don't subscribe to the social criticisms, Lena simply isn't representative of imagery nowadays. The colour balance is wrong, its from film and the dynamic range is limited.

You're better off montecarlo sampling 20 images from instagram using different geolocations.


Please please no, we'd end up having max hdr and saturation as a reference. If anything, we need to dial down these, not up.


I'm not really talking about HDR, as that a bit well, subjective.

I'm talking about using an image that that was captured by a device that can have 12-18stops of dynamic range. Something better than film, something better than 70's film rendered in magazine print and scanned in later.

or more succinctly, something that would capture the whole range of a colourspace.


I completely agree on dynamic range, and on the fact that we should use a reference image (or images) more inline with modern camera equipment. I don't believe though that _random_ instagram content would provide a good, balanced reference :-)


It's less sexist now because we removed all the women.


This is of course a joke, but most compression competitions include the size of the decompressor.


Trivially, it's effectively 0 for LenPEG 3. Answering this for LenPEG4 may be NP-complete. Any significant research into answering it for 4 is bound to drive up AMZN in the short term due to the inevitability of using all of S3's available space.

Maybe the better question is, "how long can on-disk compression protect humanity from LenPEG 4?"


> Trivially, it's effectively 0 for LenPEG 3.

I don’t think so? The decompressor needs to have a copy of the image to display if it goes to step 1 of its decompression algorithm.


I agree with you, but the other interpretation is that you could measure after it deletes itself.


One could also measure their body weight after they step off the scales but it wouldn’t be a particularly meaningful number.


If you step back on after running, you can figure out roughly how much water you need to replenish; Perhaps both numbers are useful?

When LenPEG 3 is done running, it doesn’t consume any space. Perhaps the first measurement is just considered as runtime overhead.


Worth filing under: Gaming the System, in this case: image compression benchmarks.

Not everything that counts can be counted.


You mean more like "Not everything that can be counted — counts." If everything that counts can be counted is debatable.


No, I actually mean what I wrote: Not everything that counts can be counted. Take the happiness produced by a piece of software. Can we put a number on it?

Contrast Lord Kelvin: “When you can measure what you are speaking about, and express it in numbers, you know something about it, when you cannot express it in numbers, your knowledge is of a meager and unsatisfactory kind; it may be the beginning of knowledge, but you have scarely, in your thoughts advanced to the stage of science.”

And also this doozy writ large in our lives: “If you can’t measure it, you can’t manage it.”


This is pretty much what most machine learning methods would do if you give them the task "make an image compression algorithm that compresses our test image as good as possible".


Tiny interview with Lena herself and some history: https://www.wired.com/story/finding-lena-the-patron-saint-of...


Now with the "buy a disk" modifications to lenpeg, the endgame can only be to destroy the universe during compression and do the "let there be light" if no universe is found during decompression


I'm surprised there hasn't been a solution with a docker image yet.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: