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

[JohnTF]

The school finally bought a Scan Tron grading machine, as I had used one, the boss invited people to ask me about using it.

Though the answer sheets basically have answers A-F, they have some in which you can use objective questions, but leaves a half of the page blank for essays.

Teacher next to me actually said, "Oh, the machine grades essays too?"

I'm afraid I am off her Xmas card list, I just could not quietly address that.

Right, the machine reads your student's answers, and the word recognition in a machine built to pick up No. 2 pencil marks, then interprets the answer and gives a letter grade.

I should have let her try. Too much ditto fluid I think.


On this machine, take your low scoring picture, and draw a diagonal line on it, and try again.


Regards, John

 

[Kevin]

Beauty, taste and art are the concepts encompassed by the philosophy known as "aesthetics."

I must remind all english speakers here of the important proverb which guarantees us all the individual right to exercise this philosophy.

"Beauty is in the eye of the beholder"

This software hack is attempting to consolidate billions of individual, discriminating eyes and minds into a single draconian filter for rating images.

This is clearly a cheap publicity stunt by the university and a complete waste of time and other resources.

I am extremely offended by this sinister heap of software and must declare this project to be another subversive form of hate towards mankind.

Go back to hell, Penn State!

 

[Uwe_Nds]

Thanks a lot guys, now Im depressed. Two of my photos both rated less than a 20.

Don't worry, mine rated 2/100.

Cheers,
Uwe

 

[mfunnell]

I decided to take an admittedly only semi-serious look at this, beyond the immediate causes for giggling.

The first thing I noticed is that they surround the thing with a bunch of disclaimers like:

Acquine is designed mainly to assess the aesthetic quality of color professional photographs. It is NOT designed for computer graphics, artificially-produced diagrams, figures, paintings, composite pictures, or casual family photos.

Full color photos are more suitable than black-and-white photos.

...and the big one...

As for most computer-based systems, it is possible to find special cases where the system is clearly not functioning as intended. For instance, one may find that a very poor quality photo gets a good score, or an award-winning photo gets a low score. A rule of thumb is that if the aesthetic quality of a photo is obvious to most people, it may not be worthwhile to seek Acquine's opinion on it because Acquine may assign funny scores in such cases.

Reading the published papers leading to this showed some other things:

• It likes bright, saturated, primary colours. But...
• It hates colours it sees as not complimentary (on the good old colour wheel).
• It likes the luminousity to average 18% grey across the frame (ie. doesn't like what it sees as under- or over-exposure).
• It likes a really restrictive version of "rule of thirds" composition, with a picture element (a differentiable blob, I guess) present on one of 4 nodal points.
• It hates it when a large percentage of the frame is out of focus; except...
• ...it has an exception that turns to love if there's an in-focus picture element (read: blob) in the centre of the frame. (This is a deliberate exception to the "rule of thirds" rule as they noticed that many highly-rated macro shots had a central in-focus thing and lots of OOF area.)

And lots of others, no doubt. I just picked up on a few of these rules then poked the thing with enough sticks to get a feel for how it reacts. While I got some "surprise" results (no doubt due to a combination of other rules and processing which I didn't pick up on, plus the effect of "training" it with live reactions from human ratings) I satisfied myself that I could at least partly see the results of some of these rules. (For example, by trying colour vs black-and-white versions of photos with "good" colours and "bad" colours: the colour outrated the BW in the first case, while losing out in the second.)

I suspect you could very quickly learn to "game the system" to design photos for really high (or, for that matter, low) ratings. I imagine that if this thing (or its spawn) ever gets used for anything involving money, recognition etc. then "gaming the system" could become quite important to some people...

...Mike

 

[Dave Wilkinson]

I'm surprised nobody asked it how it rates the little boy - jumping the puddle? Dave. ...... or really go for it - and show it some of Hamilton's stuff

 

[mfunnell]

I'm surprised nobody asked it how it rates the little boy - jumping the puddle?

One of our fine RFF members did that (and more):

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

HCB:

2.7

Black with noise:

88.3

Screen cap 'rating of 'black with noise" ':

51.2 (not too bad!)

Another HCB:

75.3 (better work, Henri - you may be getting the hang of things)

HCB? Bah! I have images that out-rate even his "good" one. One even out-rates "black with noise"!

...Mike

 

[JohnTF]

That shot of noise ---- though it is generally an excellent museum--- I think the Cleveland Museum of Art has its twin hanging near the entrance.

Really looks as if it were painted with a roller and some Sherwin Williams, I always wondered if the artist is still laughing his way to the bank.

I always wonder if it was part of a series, like the walls on most houses. ;-)

I do always thoughtfully put my chin in my hand, and look carefully, would not want anyone to think I am not an aesthete. ;-)

Regards, John

 

[camera.bear]

The whole idea is interesting and it certainly has a formula that it is looking for but it can't factor in the emotional impact of a photo. That said, this one of mine did pretty well.


Machine Prediction of Aesthetic Value [SIZE=-2]

:[/SIZE]99.7 / 100

 

[paul beard]

1.8 / 100 on this.

Can anyone top (bottom?) that?

And 25 / 100 in this:

I think they are trying to backfill an emotional response element by asking for your assessment (the 1-7 stars). But without knowing what's in the frame, how can that work? Are they working toward whatever non-random arrangement of pixels is Art, as defined by combining the two metrics? Good luck . . .

 

[amateriat]

So, Penn State, essentially, managed to create their own, small-scale version of Deep Thought.

I'm all for Pure Research and stuff-like-that-there, but how slow a semester must that have been?


- Barrett

 

[Morca007]

Much closer to Deep Thought would be the soon-to-launch Wolfram Alpha

 

[Sparrow]

One can see where it's sympathies lie ... i keep warning but no one else thinks the machines are coming

 

[varjag]

The computer does not analyse aesthetics. It's just one of machine learning algorithms trained over massive set of pictures with their popular ratings. It perfectly reproduces mediocrity and lack of taste as seen on numerous photography-related websites.

Don't blame the machine

 

[charjohncarter]

When your now Senator (ugh) was mayoress (around 1985) of San Francisco, CA, she had at least one good policy: 'No new computers until I see some productivity increases.' None were purchased, at least for a while.