Project 3; Chair Genealogy
The idea behind the project was to build a “tool” of some sorts to generate hodgepodge mash-up chair designs from mixing and matching pieces from prominent Contemporary and Modern furniture designs. My hope was to create a unique design from these pre-existing ones and to pair their parts in ways that may not usually be considered.
I used Dan Shiffman’s example of a genetic algorithm paired with user input from his book “Nature of Code” to make this project. The program stores different attributes about the parts as well as the parts themselves as genes, which are then carried over or phased out in the next generation depending on the “fitness” of the combination. (The fitness is the number displayed below each chair option.) There is also a mutation function to keep a bit of variety by adding varying degrees of randomness in the gene pool each time a new generation is produced. By pressing ‘s’ and image of the designs currently displayed on the screen can be saved in the project sketch folder.
There are a few things I would like to improve about the program such as having a more refined method of aligning the parts, more selective way to save the images, to push the bounds of what constitutes a chair and eventually, I would like to make some of the designs produced by the program.
————————–
Its funny that it is 2D. – why?: Because it has less application as a tool for having skematics or sketeches for a builder to create the furniture, but is more a comment on design magazines documentation of chairs. WHICH IS NOT A BAD THING. Just def a difference between builders and documentation.
^ sure, but when you write “its funny that it is 2D” you need to explain yourself
You can increase fitness by hovering over, but how do you decrease fitness?
Artists historically have been using chairs as a stand in metaphor for the human form. Arthur Ganson did this many times with his kinetic sculpture that could change and rebuild over time. I think it is interesting in imagining how our bodies could relate to these new forms of furniture.
Classic font choice. Goes well with the modern theme of your project
Why chairs? I’d like to see some of the chair results and why you chose them. Perhaps show some interesting examples before the functional explanation, results of your exploration.
^^I feel like you should explain why you chose chairs.
It might also be interesting (though not as pretty) to generate meshes with some kind of physics sim in order to maybe evolve some really crazy chairs verses combinations of ones that already exist, especially if you are planning to use the app to generate ideas for actual furniture. Like you could have any number of legs of any lengths or arms or strange curves that might lead to some interesting results. Also, how do you determine which is actually most “fit”, since there is only the images?
I think using additional images would help make more randomness; also, maybe having more components than just top/bottom… like also back, base, arms, headrest, etc… would create more possibilities. Also, allowing any image to be used for any piece, though it would generate a lot of duds, might also generate more possibilities (oh, you just explained this)
^^ I just think she hasn’t cycled through all of her chair parts yet.
I’m surprised that just splicing images together generated reasonable looking results (for the most part). I wouldn’t have thought of that approach.+1
How are the fitness ratings decided? Is it just based on the amount of time your mouse is over a chair? That’s what it looks like from the demo, but it seems like a very imprecise input method.
i’d be interested to hear if other designers find this useful cool idea
Cool idea: evolving furniture. I would have loved to see a finer granularity on the chair pieces. It mostly looks like the chairs have a top & bottom (although that may just be this species that is on screen at the moment). Mixing and matching chair legs for a single chair might result in some fun images. The voting mechanism is also a great and simple way to evaluate fitness. I’d also love to hear some hypotheses about algorithms for evaluating fitness based on some parameters, even if you couldn’t implement in the given amount of time.
————————-
I think you definitely learned about the process of how to deal with such algorithms and no doubt it took you a while to figure out. However, I wish that the chairs could have been morphed together to form a solid form instead of a ‘cut & paste’ sort of deal or there could have been another way of determining fitness than a mouse hover. Regardless, I think this was a very informative experiment.
I wonder how difficult it would be to break the system, so that all chairs look the same.
Are there only two “pieces” for each chair? I would’ve liked to see at least back / base / seat/ arms.
^^Essentially you dope one parent/style’s fitness score and hope it doesn’t mutate
I really like this idea specifically looking at cross genres for furniture styles and discovering new styles. I think the fitness rating might be better as a priority (like rank your favorites) maybe a like button instead of a number.
Very cool. Check this out: http://sketchchair.cc/. You may be able to combine the idea of your chair generation with automatically producing a fabrication file. Crazy custom.
As an idea-generator — this is FANTASTIC.
Pity that the chairs are so small on the screen.
Please make sure to mention where the parts come from, and how many there are.
Have you worked out the total size of the combinatoric space? e.g. #of arms X #of legs X # of seats..= 40,000?
I would have liked to see a simple visual display (like a bar) instead of a number, by which the user specifies the fitness of each chair… just as a better form of interface design… but this is a minor point :)
The chair combinations might show more evolution given more parts
love the aesthetic. very ikea-esq with the white bg and font choice. very appropriate! :)
the concept is a nice twist on the theory of evolutio, etc. can see this layout being utilized for many different image databases.
I love your concept but I agree with the comments that your interaction design could use some further work. I’d be great to see the history of your chairs, the ‘family tree’ that lets you visualize the genealogy. Your presentation could benefit from some awesome examples, or maybe a walkthrough of a scenario where you keep picking the same legs and see those legs propegate over time. Right now its pretty abstract, its up to us to infer how this actually works.
Interesting idea! I like the designes of chairs (they are so modern), but it is true that you can develop this generative algorithm more to create something really new that viewers cannot expect. All these chairs are kinda expectable. Maybe you can morph two different chairs into a new form.?
Golan’s suggestion could maybe be implements as Affine Transforms on the image with coordinates described by the algorithm
————————-
I love your wallpaper :) lolz hahaha
Great UI
Cool way of creating iterations of combinations of design elements. I sorta wish it took up the whole screen.
I wonder if you could get more parts from some online source- like a furniture parts retailer.
For exmaple- sofa legs (http://www.furniturepartsonline.com/sofa-legs.html) and wheels (http://www.furniturepartsonline.com/wheels.html)
If you could explain the algorithm a little, I am curious about that.
** Agree **How is the fittness score calculated? **fitness scored is assigned by the user, it is not calculated. It is then used to decide for each generation, who is “more fit”, and thus its characteristics should
be represented more (if that makes sense).
Beautifully laid out on the screen. I would have chosen a bit of a smaller font for the headline “Chair Genealogy”.
Maybe can you show the chairs a bit bigger? I would love to see a bit more details.
Add a label that says that the numbers represents fitness values. It will help the users understand what should they do.
If there are more componets of the furniture, and there will be a more interesting and complex combination of them…..
** agree.
Simple and nice interface!
The design is very clean and simple, I really like the chair images!
** I agree.
It’s very cool. And a fun design tool. I would like to see a number of iterations generated rapidly to see what kind of variation occurs.
————————-