Pages: 1 [2]   Go Down

Author Topic: FWIW Artificial Intelligence Critique  (Read 4383 times)

Arlen

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1707
Re: FWIW Artificial Intelligence Critique
« Reply #20 on: October 18, 2016, 09:40:23 pm »

Though it's clearly questionable from the get-go that an algorithm like Keegan can accurately determine which images are good (defining and measuring "good" being the key problem), it seems more feasible that it might be able to predict which ones will be popular.

As a limited, non-rigorous attempt to test that idea, I went back through several several months of images posted in the LuLa Critiques and Landscape & Nature forums. I fed about 60 images to Keegan that were popular in those forums, as indicated by larger than average numbers of positive comments.

I could see no consistent correlation between Keegan's ratings and the apparent ratings conferred by LuLa reviewers.

But I did notice that Keegan does not seem to favor complexity, low contrast, subtlety, muted colors--or "street" photos of almost any sort. For many abstracts, it strangely complains that it can't analyze them at all because they are not photographs. On the other hand, it does seem to be rather fond of well-lit, high contrast, boldly colored, sharp, simple and large main subjects.

Take home lesson: it is easier to divine the image preferences of Keegan than those of LuLa reviewers.  :)
Logged

Eric Myrvaagnes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 22814
  • http://myrvaagnes.com
    • http://myrvaagnes.com
Re: FWIW Artificial Intelligence Critique
« Reply #21 on: October 19, 2016, 12:01:37 am »

So even a computer algorithm can be biased. Maybe it was abused as a child.
Logged
-Eric Myrvaagnes (visit my website: http://myrvaagnes.com)

Rory

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 528
    • Recent images
Re: FWIW Artificial Intelligence Critique
« Reply #22 on: October 19, 2016, 12:10:19 am »

I can confirm it likes birds, especially in flight.  Also dragonflies, which it seems to know are not birds.  10 bird images scored between 8.8 and 9.0.  Anything abstract not so much.  :*)

♫I'm like a bird, I only fly away♫ oh god! Don't sneak up on me like that. Well placed subject, good job on the lighting, and blurry but elegant. Thank you for this interesting image and very good timing; I'm proud of you. It's simple and it's really good. 9.0/10 - you deserve it, champ.


Logged
[url=http://www.flickr.com/photos/roryhi

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: FWIW Artificial Intelligence Critique
« Reply #23 on: October 19, 2016, 01:40:50 am »

Hi,

I think it uses rules of composition to a great extent, also analyses light and looks at cropping. Depth of field and bokeh also matter. Obviously it considers a lot of other factors.

If you are at a camera club and discuss pictures there are three ways:

1) Either you like a picture or you don't
2) Strikeout, compare two images and drop the looser
3) Talk about rules of composition and light

What "Keegan" does is probably to a great part (3). I have never really been in favour of that stuff, following rules doesn't make a good picture, but many good pictures follows those rules. So an image with good rating is probably well composed and in decent light.

I don't think it understands fun.

Best regards
Erik


Though it's clearly questionable from the get-go that an algorithm like Keegan can accurately determine which images are good (defining and measuring "good" being the key problem), it seems more feasible that it might be able to predict which ones will be popular.

As a limited, non-rigorous attempt to test that idea, I went back through several several months of images posted in the LuLa Critiques and Landscape & Nature forums. I fed about 60 images to Keegan that were popular in those forums, as indicated by larger than average numbers of positive comments.

I could see no consistent correlation between Keegan's ratings and the apparent ratings conferred by LuLa reviewers.

But I did notice that Keegan does not seem to favor complexity, low contrast, subtlety, muted colors--or "street" photos of almost any sort. For many abstracts, it strangely complains that it can't analyze them at all because they are not photographs. On the other hand, it does seem to be rather fond of well-lit, high contrast, boldly colored, sharp, simple and large main subjects.

Take home lesson: it is easier to divine the image preferences of Keegan than those of LuLa reviewers.  :)
Logged
Erik Kaffehr
 

Arlen

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1707
Re: FWIW Artificial Intelligence Critique
« Reply #24 on: October 19, 2016, 02:13:30 am »

Hi,

I think it uses rules of composition to a great extent, also analyses light and looks at cropping. Depth of field and bokeh also matter. Obviously it considers a lot of other factors.

If you are at a camera club and discuss pictures there are three ways:

1) Either you like a picture or you don't
2) Strikeout, compare two images and drop the looser
3) Talk about rules of composition and light

What "Keegan" does is probably to a great part (3).
I have never really been in favour of that stuff, following rules doesn't make a good picture, but many good pictures follows those rules. So an image with good rating is probably well composed and in decent light.

I don't think it understands fun.

Best regards
Erik


I agree that's probably what it does, and most of the images it assigns high ratings are ones that many of us would say are at least reasonably OK, and sometimes very good. So it's not useless, and is actually kind of fun. Hey, it's early AI days, so how much can we expect at this point? But among the majority of images that are assigned lower ratings are the very ones that many of us would say are the very best. Of course, we won't agree among ourselves which ones they are.
Logged

David Eckels

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3526
  • It's just a camera.
    • Website
Re: FWIW Artificial Intelligence Critique
« Reply #25 on: October 19, 2016, 07:55:48 pm »

But do I really?
Quote
"Really elegant perspective and the lighting, yass. You got it, chap! Congrats!! You get a 8.5/10."
#perspective#bright

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: FWIW Artificial Intelligence Critique
« Reply #26 on: October 22, 2016, 04:25:58 am »

Some examples…

BR Erik
Logged
Erik Kaffehr
 

Eric Myrvaagnes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 22814
  • http://myrvaagnes.com
    • http://myrvaagnes.com
Re: FWIW Artificial Intelligence Critique
« Reply #27 on: October 22, 2016, 07:59:01 am »

Thanks, Erik.
So we can conclude that cats are better subjects than landscapes or closeups.   ;)
Logged
-Eric Myrvaagnes (visit my website: http://myrvaagnes.com)

Rob C

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 24074
Re: FWIW Artificial Intelligence Critique
« Reply #28 on: October 22, 2016, 08:50:06 am »

Thanks, Erik.
So we can conclude that cats are better subjects than landscapes or closeups.   ;)


Dd anyone try a nood?

Rob

GrahamBy

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1813
    • Some of my photos
Re: FWIW Artificial Intelligence Critique
« Reply #29 on: October 25, 2016, 07:49:07 am »

In other AI news, the completely AI generated pop-song. It's impressive from one point of view, but a certain amount of nuance remains to be mastered. Then again, it may be comparable to efforts by eg Japanese bands to write English language pop songs...

https://www.youtube.com/watch?v=LSHZ_b05W7o

Logged
Pages: 1 [2]   Go Up