Friday, 9 May 2014

One thousand works of Systems Art - how and why


In my my previous post I published a video entitled One Thousand Works of Systems Art. I want now to explain my motives and my methods in constructing those 1000 images and that video.

The panel above shows the images at a more leisurely pace, where the images play in random order, one every three seconds.

The video runs through exactly 1000 images at the rate of 4 per second, so the whole thing is complete in a little over 4 minutes,  [if you haven't already, you can see video on Vimeo].

If you watch the video, you will see some logical order to the images.The same themes re-occur and consecutive images seem rather similar. The video has a music track which comprises 1000 chords synchronized to the video frame rate. The work is intended to be viewed both with and without the music. All this will be explained in what follows.

My motivation in preparing the video was to see how systematic I could be in presenting similar but different images by generation.

I have made some mistakes. The images are too similar. But this is a first experiment and can be considered a proof-of-concept. I need to define what I mean by "sufficiently different", before I take this further.

The method of generation used, which should be obvious, is collage. The composite images are made by pasting sections of photographs onto a background. Some of the photographs are simply of brushstrokes, which give the collage a painterly effect. Others are sections of scenes from life but with an abstract character (e.g. a corner of a building, not obvious until you've looked at it for a while).

The choice of fragments to paste into the final image is random-ish. It's not completely random, because I need to avoid the situation which may occur when the choices conspire to effect the same result. So it's a controlled randomness. A bit Dada, perhaps.

All the resulting 1000 images, if they were to be classified as anything, could be called Abstract Expressionism, or Lyrical Abstraction. At least, that was my hope. Channelling Clyfford Still, or Jackson Pollock or early Rothko.

Although it won't be apparent, the sequence does have emotional content for me, in that all the component parts were collected by me on a long, happy Holiday in Italy.

But there are technical issues to address. The fact that I can be confident that all the images are unique is based on a mathematical argument that guarantees that, even though the components are chosen at random, no collage contains exactly the same components as any other [not quite true, as it turns out, see below]*.

The method of generation I call Compositional Form. Each final image is a composition of component images. The program that generates the 1000 final images is a "script" that first selects a number of component images and then composes them according to a predetermined formula.

It you are interested in the more technical details, there is a short note at the end of this post which describes the programming ("scripting") involved.

This may or may not be art. It combines artistic elements in a way that uses both elementary Maths and simple Computer Programming.

The mathematics enables me to calculate registration for images of different sizes and to determine the transparent regions of a cutout.

The scripting allows me to experiment with different sets of components and with different definitions for my composing operations with the greatest of ease. I can change the script and rerun it to see if the artistic aspirations have been more closely reached.This is well nigh impossible to do by direct manipulation in a system for manipulating images, because of the large number of identical operations that need to be repeated. The operations need to be scripted so they can be repeated 1000 times.

The way I have chosen to display a complex artwork like this is video, which  in my opinion is well suited to the task. Combining the images into a satisfactory sequence also involves a little maths and a little scripting. This has mainly to do with guaranteeing the duration of the video and with synchronising the music. The music chosen for this video (if you can call it music) is running at 120 bpm, which is 2 bps but there are two notes to each beat, hence 4 notes per second. Exactly in sync with the images at 4 per second.

The music was generated using a Compositional Form similar to that for the images, but rather more complex. I will go in to that in a later post on this blog.

My ambition for this idea is to generate a sequence of 1000 images that are sufficiently different in their form that you would not particularly consider on to be a close copy of any other, but to make the sequence in which they appear to be sufficiently interesting that someone would tolerate watching them for four minutes.

------------- Warning - the remaining notes are a bit technical

If the component images in any work are, for example, image1, image2 and image3 then the composing formula might be

   overlay(cutout(image3), overlay(cutout(image2), image1)

where cutout(image) takes a image and trims it (as if with a pair of scissors) to be something that can be pasted onto another image covering some of it and allowing some of it to show through. The composing operation overlay(image,image) simply composes two images to form one, with the first on top of the second.

Thus the simple formula we gave above reads "a cutout from image 3 overlaid on a background, which itself is a cutout from image2 overlaid on a background which is image1".

Since the operations (overlay and cutout) are turning images into images they can be composed in quite arbitrary and quite complex ways. This is where I expect the "sufficiently different" variety to come from. For example, it is quite possible to make a cutout from an image that has been constructed from an earlier cutout -

    cutout(overlay(cutout(image2), image1))

where we have first constructed an image which is a cutout from image2 overlaid onto image1 and then made a cutout from that, presumably to stick onto something else.

The library of operations available to me for Compositional Form includes a small number of additional operations for producing interesting new components. For example the operation waste(image) is the opposite of cutout(image) in that it constructs the part of the image that would have gone to waste (fallen to the floor) had we actually been using scissors.

The fact that the 1000 images have been generated in this way is apparent from the sequence you see in the video. Adjacent images are strangely similar. There are runs of 10 or so images where some of the component parts do not change (although, of course, some do).
-------------------

*[Because I accidentally included the same component image twice in the collection of components from which the collages were made, it turns out 10 of the images are identical. They are not adjacent in the video sequence, so it is not apparent. The mathematical argument which ensures uniqueness of the generated images is not affected, for it presupposes that all the component parts are different. If I want to guarantee to someone who takes possession of a (high resolution) image that they have a unique copy that I will not allow anyone else to have, then I need to ensure that all of the component parts are indeed different]