Say goodbye to criticism, technology is here to rescue us!

[climbing_vine]

Someone show it some Noctilux photos at f1 and see if it appreciates premium bokeh and vignetting! 😛

Keith, if they're smart they're checking the EXIF info and giving Leica shots 100 so as to spare themselves from the last 23 Leica cultists in the world shutting down their mail server with flaming outrage.

 

[climbing_vine]

No computer will ever 'understand' abstract concepts like love, fear and anger so why bother asking one to judge photos? Great photos have the potential to say so many things on so many different levels while others may not say anything at all, but may just be nice to look at.

I wonder how some of Rembrandt's images score.

Why haven't they made a computer that judge beauty contests?

You seem to have missed the part where it's already pretty accurate and getting better every second.

You only have to read the average politics forum on the internet to find that what most people "understand" about love, fear, and anger could easily be distilled into less than 100 lines of code in Perl (thanks to its convenient regular expression parsing).

There are outliers, but they're exactly that.

 

[Keith]

This sotware, although it obviously ignores and can't comprehend genuine artistic merit, would be very adept at picking images for post cards, biscuit tin lids etc etc!

So it does have some use in this plastic world we live in! 😛

 

[ferider]

Highest rated pics on photo.net:

http://acquine.alipr.com/toprated.php?sidx=0&ord=0

It's for color, 600x600 or bigger images. It just went on-line (in April). Its performance is not limited by the pics it's being fed, but by the user rating them. So keep going 🙂

Cheers,

Roland.

 

[FallisPhoto]

Some of the junk I never even bothered doing anything with scored 90+. One of my prize winners scored 58. This is a joke.

 

[climbing_vine]

I really don't think ten thousand years of human evolution, religious and philosophical thinking and art could be 'understood' by any computer.

I wonder if a computer would have an 'understanding' of what this little boy and his two friend's lives are like. They live in a remote village in the Highlands of Papua New Guinea. They have no electricity, no TV, no radio and believe that white people are ghosts.

I won't say "you probably don't either" because I don't know you, but the law of averages would make it a justified statement.

Virtually none of the industrial world has any understanding of this boy either before or after seeing this image. In some people it may elicit a brief (very brief) emotional reaction, mostly because in this particular image he looks "lonely" in a scene with no other people.

I think you have a very high opinion of the "understanding" that an image supposedly imparts, which is self-evidently delusional. If images like these gave anyone "understanding", we'd all be giving enough time and money to have eradicated hunger in the third world.

I'm sorry, but these are the facts.

 

[Bob Michaels]

correlation of on-line ratings vs. personal perception of merit?

correlation of on-line ratings vs. personal perception of merit?

I read the write up of this system on the Penn State website. It does seem to be honing in on some predictor of user ratings that one would derive from traditional on-line photo review sites.

But, what correlation do others see between their personal perception of meaningful photos and the typical on-line ratings?

I will open by saying I see minimal correlation in ratings and my perception. It seems to me that standard forumlaic photos get rated highly, probably because that is the standard the reviewers are trying to reach. Much differentiation from the norm seems to be the kiss of death in user ratings. However that is just the way I see it.

How do you see it?

 

[ferider]

Much differentiation from the norm seems to be the kiss of death in user ratings. However that is just the way I see it.

How do you see it?

Exactly like this. It learns about "the norm". In aesthetics, not semantics.

The stupid thing just gave Da Vinci's Mona Lisa a 15.5.

Do you think the "average" person finds the Mona Lisa "prettier" than one of the highly rated nudes shown on photo.net ?

Roland.

 

[samoksner]

This shot by a photo god himself: HCB, got a 2.7! Not joking.

http://www.magnumphotos.com/CoreXDoc/MAG/Media/TR3/F/W/Z/E/PAR44919.jpg

Try it yourself to see what it gets...

 

[samoksner]

I just showed this thing a picture of a black wall with some digital noise, it scored an 88.3.

Here is the image that fared so well: way better then HCB for sure... http://gallery.artofgregmartin.com/tuts_arts/stars_images/01.jpg

 

[Keith]

'black wall with digital noise' ... love it!

Where can I get a copy? 😛

 

[Al Kaplan]

Bob, the way I see it is that we have to get to the point where we don't much care what other people think about our photographs. I spent too many years shooting to meet clients' expectations. Bills had to get paid. Now I just shoot what I want when I want. What's the point in attempting to shoot photos that will get high scores from a machine? It's just carrying playing video games to another level because it reduces photography to just a game.

 

[climbing_vine]

Korda's image of Che scored 56.8. That's a D- on a school report. This image, as we all know, is the most reproduced in history and probably one of the most emotive, yet simply composed images ever. The man, the myth, his politics, the misuse of the image, even current politics play their part in how we view this image. What machine could ever 'hope' to rate it?

It's explicitly rating aesthetics, not socio-political significance. I continue to be amused that people are ignoring the stated point of the thing. Again, says a lot about the supposed "perception" that is so highly valued.

 

[climbing_vine]

There you go Brian. You just proved my point. Your 'very brief' emotional response in attributing loneliness to the little boy is beyond the realms of any computer. So why bother asking one to judge photos?

I must apologise for my obviously self-evident delusional thinking. I'm only human.

I can tell that you're not stupid, you're just deliberately misrepresenting what I'm saying. Which is, indeed, human and makes you not worth trying to have a normal discussion with.

 

[Bob Michaels]

Bob, the way I see it is that we have to get to the point where we don't much care what other people think about our photographs. I spent too many years shooting to meet clients' expectations. Bills had to get paid. Now I just shoot what I want when I want. What's the point in attempting to shoot photos that will get high scores from a machine? It's just carrying playing video games to another level because it reduces photography to just a game.

Al, you and I are thinking alike here. I quite posting photos for comments several years ago because I felt I was letting the responses subconsciously push me toward the middle of the road. Similarly I have refused to sell prints although I have been somewhat generous to many who showed sincere appreciation.

 

[climbing_vine]

I read the write up of this system on the Penn State website. It does seem to be honing in on some predictor of user ratings that one would derive from traditional on-line photo review sites.

But, what correlation do others see between their personal perception of meaningful photos and the typical on-line ratings?

I will open by saying I see minimal correlation in ratings and my perception. It seems to me that standard forumlaic photos get rated highly, probably because that is the standard the reviewers are trying to reach. Much differentiation from the norm seems to be the kiss of death in user ratings. However that is just the way I see it.

How do you see it?

I think you're exactly right. You're part of a totally insignificant minority (speaking in the larger scheme of things) who take images seriously and (sometimes) think about them. I'd like to think I am too, but depressingly I have to admit that my tastes tend towards the mass ratings (I think most of HCB's work is vapid and dull, but I love the pretty colors in Cezánne).

A lot of people are purposefully ignoring two things. One, this thing is explicitly rating aesthetics, not socio-historical significance; and two, it's a lot more accurate at guessing what will get rated highly on photo.net than any of us are.

And I just have to add that the Che example used by someone else is a perfect one, but not in the way they intended. It is the most-reproduced image in the world, but 95% of the people who wear it or hang it on their wall know nothing of its significance other than "my friends in the fraternity told me he stands for something about the freedom to get high". The idea that it's a pervasively *affecting* image is laughable, and demonstrably false. If 1/10 of the people who wear the shirt actually knew and shared his beliefs, capitalism would have fallen by now. 😛

 

[climbing_vine]

Socio-political significance forms part of my aesthetics. Just as religion plays a part to some people when viewing religious images. Some images have meaning to some people which is beyond the comprehension of any machine. As you just proved by your response to Korda's Che.

What exactly does this mean: "which is, indeed, human and makes you not worth trying to have a normal discussion with."

Are you a machine?

1. You're abusing the term "aesthetics".
2. If the machine were as bad at judging all of this as you keep insisting it must be, it wouldn't... um... be as accurate as it is.

QED. I'm not quite sure what your point is, other than being argumentative.

 

[ampguy]

It is working!

It is working!

I like the 2nd shot the best. The software works!

The algorithm appears to love my bridge photos.

91.9

And likes the same bridge shot on film slightly better than digital, despite vignetting.

96.9

On the other hand it doesn't like classic English cars shot on Velvia.

13

It hates American blues drummers in Europe shot on digital.

6.8

I don't think I buy even the modest claim that this system is doing much of a job predicting mass appeal. We are part of the masses too, and should be able to see some kind of visual argument for these ratings, but I'm just not getting it.

 

[ampguy]

More proof that it works

More proof that it works

This portrait of an RFF member, was taken with a D40x, and 105/2.5 PC Nikkor lens. It scored 93/100, and of the 5 snaps, I would agree that this one is the best of the set. It is refreshing to know that it does not bias against blown highlights (which were intentional, of course ...). Also interesting to note that it does not discriminate between original JPG size (~3MB) vs reduced 640 pixel on long side < 200Kb sizes.

 

[notturtle]

computers are no better at jokes