Continuing on from yesterday, further tests and developments were made to the sculpture prototype. Ana spent most of the day adjusting the 3d model in Blender.

Further adaptations were made to the painting software. This software is a development of a patch that was used as part of the New Fordist Prototypes performance at the beginning of the year. It can be seen at work below:

Using the data we gathered last week, I attempted to find lines of best fit for the relationship between time taken and brush-stroke length.

Fig. 1 Line Of Best-Fit On Scatter Graph of Brush Stroke Length / Time Taken To Paint for Brush Size 12

These were calculated for each brush size surveyed, creating a set of five exponential equations:

These figures were then averaged together to smooth out the anomalies in the data and give an equation that could be used to estimate the time that a brush-stroke of any length would take to paint, across all brush sizes (it is hoped at a later date to more accurately calculate this using further data and a variable that accounts for brush size, but, as an experiment, this serves the current purpose).

Averaging the two fixed variables in the equations, gave a new equation of:

f(x) = 3919.012 x exp(0.00398921x)

This equation was then used in the painting software to calculate the time that a brush-stroke indication should be kept on-screen for.

The patch now uses this data to calculate the amount of time the picture will take to paint and, using the information on paint consumption from last week, the amount of paint needed to paint it.

Added to this, the patch now looks for colours between two extremes, rather than a specific hue. This addition will allow the overlapping of different colours so that layers can be created - a technique closer to how painting would normally be done without the program, which had previously isolated the colours in works containing more than two colours.

To test out these improvements, another version of the portrait of gramsci, used in the video above was created - this time using the original photo as the source, rather than a colour-reduced image. This time, the aim is to use two colours with over-lapping colour ranges, the new equation for the timing and check that the estimations provided by the program prove correct.

First, the colour range that the software searches for was inputed. The aim is to do a mid-range grey as a base layer, and then paint the, much less voluminous, darker sections on top. The brush size chosen was 12 and the colour range was between 0 0 0 and 160 160 160 RGB. The mid-range colour goes all the way down to darkest black so that it provides a base layer for the black, when it is painted on top.

Second, the program analyzes the image, a process which took 11mins 42 secs (this image was 400x300 pixels). During this procedure the computer calculates where each brush-stroke should be on the image and stores this in a table to be read back later.

Third, the estimations are noted down, for checking at the end. The computer estimated that:
1. This layer would take 40mins 44secs to paint
2. that 31.13ml of paint would be needed.

Fourth, the painting commences - below is the process of painting, photographed at one minute intervals:

Fifth, an analysis of the estimations:
1. The painting took exactly 40mins 44 secs to paint. However, the speed should be increased as there was still some time between brush strokes - this will be done by altering the fixed variables in the equation.
2. The paint estimation was wrong - in fact, rather than 31.13ml of paint used, in fact 45ml of paint were used. This underestimation means that this calculation will have to be re-assessed.

Unfortunately, there was not enough time to do the second layer today and we will not be back in GEMAK until saturday. However, new estimations were calculated for the next layer (the analysis for this layer took 11mins 44 seconds:

1. At the current speed, it will take 60min 6seconds. However, I will increase the speed for saturday.
2. 28.62ml will be used.

We shall see on saturday if this is correct.