MONDAY
Some ideas:
- histogram-to-3Dmodel
- cropping (reduces complexity at the most complex part of the model)
- stretching
- export the diff to a 3D model
- A music box, the image as a recording
- look at morphing shapes in inkscape (inkscape needs to decide on an order)
- morphing objects with same amounts of faces
- to have the pixel representation follow the rotation of the model
- what would the representation of a sphere look?
- Could you compare different versions of Blenderella?
FOR TOMORROW?
- modify the fileformat
- what you can do with the output
wii pad ... measuring that Julien's sister gained weight since she played last
The difference that is difficult/mportant/significant
Where is the activity of the model? Differences between documenting the action and conclusions on the behaviour.
Is it subjective, where the activity is? ("I would spend hours on picking a typeface, but make lay-outs very easily"
Same amount of faces ... does it need to be constant? How do you animate this in Blender?
Blender can currently only interpolate (using shape keys) between shapes only if they have the same amount of faces. How can this be overcome? It is easier to do in 2D such as Inkscape and Synfig where the node count can be different.
It makes a difference if the diff maps "a tear" or "a morph" (a topological change)
Make Human Quotes:
"By resetting your human you will lose all your changes"
Antonio quotes:
"Make human is obviously an export tool"
"Off with his head"
"Oh, I am just glitching the file"
Phil-quotes:
"Political and technical (?) diffing"
"Fidelity is resolution"
"You cannot go from a sphere to a circle"
"5 megabytes ,,, that is a lot of faces!"
"That was surprisingly disappointing, wasn't it"
"I don't know whether it is good or bad it is just a thing"
"t was not a design feature, it was just an outcome"
"The more faces it has, the more difficult it gets"
"At some point the pixel would not be read as geometry"
"You cannot do a diff unless you have a structure"
"Stick THAT in your RGB cloud!"
"It's hip to be square"
Bibliography
Simulation and Its Discontents, Sherry Turkle
Interrogating the Data Model:
A 3D model (STL): a closed shape of triangular faces.
Every face has 3 vertices.
Each vertex has 3 coordinates (x, y, z)
Each coordinate is represented by a colored pixel.
A pixel is discribed by three color values (r, g, b).
Each coordinate is mapped to color values going from 0 to 199. (Color values beyond 200 are reserved for other types of data)
Color values under 100 represent negative coordinate values. (And so color values above 100, represent positive coordinate values).
TUESDAY
the colour is just an artifact
DATA Model version "saucisson"
255 slices. (Low resolution)
Per slice: 1 pixel for X, 1 pixel for Y. (High resolution)
Take the value of X (or Y), multiply it by 1000 before casting it to an integer, and use that value for the ARGB color of the pixel.
As a result, each line of pixel should be a sequence of X and Y pixels representing vertices in the corresponding Z slice.
WEDNESDAY
DATA model mods. (Going full color)
Going full color resolution. It was decided to find the bounding box of the object and map that bounding box to the full color scale available in a 2D image.
After some experiments with 2D tools, the main factor that would modifiy a model would be transparency changes.
So, for other testing purposes, the data model was changed to a full color model without transparency. Where the modifications in the reds would have the highest impact and the modifications in the blue the lowest impact.
FRIDAY
How to convert PNG to a movie on the command line:
sudo apt-get install melt
1 image per second (25 fps):
melt -group out=25 vlcsnap-2013-12-14-12h*.png -consumer avformat:slides.mp4
5 images per second:
melt -group out=5 vlcsnap-2013-12-14-12h*.png -consumer avformat:slides.mp4