Hyperion wrote: ↑Fri Dec 01, 2017 10:48 pm
What do the collision boxes for these shapes look like? my main concern here is that visually there will be empty space, but you slam into the invisible wall of a mismatched collision box.
Collision boxes will be as accurate as we can afford to the shape of the object. So no, you won't collide with huge amounts of empty space XD
Hyperion wrote: ↑Fri Dec 01, 2017 10:48 pm
Can these warp functions be ran in an arbitrary sequence an arbitrary number of times? if so could I request a torus that's been extruded, beveled, extruded, beveled, extruded, axial pushed, greebled, and then beveled one last time
and if it looks pretty cool, could you make a cluster of them?
Yes, I'll play around with that request on Monday and see what I can do! XD
Hyperion wrote: ↑Fri Dec 01, 2017 10:48 pm
Have you thought about having concentric frame tessellations?
You're going to have to link me a picture.
Cornflakes_91 wrote: ↑Sat Dec 02, 2017 11:09 am
Do a check pass and merge the vertices of edges with length < cutoff?
Would remove the weird polies and would prevent any weirdness caused by the ignored polies
I have thought about solutions for removing/mending bad polys, but it would be expensive to run the more complicated the mesh got. I'll work on implementing a solution if ignoring them turns out to be not good enough. For now, even super-complicated meshes don't seem to have any quality taken away by my current approach.
Cornflakes_91 wrote: ↑Sat Dec 02, 2017 11:09 am
Well, thar archimatrix stuff forgothisname posted in the other devlog thread had some interesting concepts that may be usable for that end.
You can create prefabs which also encapsulate rules on placement.
So if you can get a similar system running it would be very beneficial for such ends.
At least for detail/surface elements.
Take object -> apply "style" elements.
Maybe warps and transformations can be packaged and reused as well in a similar fashion?
A different thought on the style thing would be to generate models with some parametric sheme and with enough parameters left free to change them into a large variety of ship/-classes.
So the engine doesnt have any clue about the style, but still creates similarily styled ships by varying the parameters.
Both shemes require a powerful parametric design framework which is probably a big obstacle in itself
Also relevant
viewtopic.php?f=6&t=3259
You've sent me that video before, I quite like it!
Currently, yes, warps and transformations also have the ability to be packaged and re-used with Style, along with shapes and other generation constructs. The problem isn't that I don't have ideas about how Style can be applied; it's that I don't have a solid enough ship or station-building algorithm to test them on yet.
Victor Tombs wrote: ↑Sat Dec 02, 2017 3:02 am
Damocles wrote: ↑Sat Dec 02, 2017 2:40 am
BTW, you should keep hold on those concept drawings for subdevisions.
Thanks for the memory jog, Damocles.
Any chance that you will get your own section in the book "Architecture of a Universe", Lindsey? And can we high tier backers get one of your drawings to accompany the scribblings of Josh? Although it will be awesome to receive a page from Josh it would be equally as awesome to receive one of your concept drawings.
I'd push my luck and ask if François could include a sheet of music from his score for Limit Theory but perhaps that's a step too far.
I intend to cram as much of my own work into the book as possible! ;D
Also LOL @ Damocles, I'd have to be a particular kind of famous for THOSE scribbles to end up selling for anything XD
Victor Tombs wrote: ↑Fri Dec 01, 2017 11:20 pm
Is it possible to confirm that stations and ships will eventually have lights and "windows"?
Yes on lights; no on windows.
mcsven wrote: ↑Sat Dec 02, 2017 1:46 pm
Great stuff Lindsey, thanks for taking the time to update us and provide shinies. It's great to see a team working on LT; I can't imagine what a relief it is for Josh.
My perception is that problems like this - which are particularly difficult to hand-code - are generally solved by machine learning these days. Have you guys considered, now that you have the basic shapes and transforms, knocking together a ship editor and releasing it? Then we can all play around and submit our ships back to you guys so that you can train a machine learning algorithm?
It would offer another solution to the problem in addition to whatever hand-tuned algorithms the dev team come up with.
MP X10L wrote: ↑Mon Oct 30, 2017 8:16 pm
I wonder, when it comes to recognising shapes ... aren't neural networks pretty good at that?
You could feed your analyzed stations and doodles into something like hyperGAN. It spits out images of Stations shapes that it thinks are pretty. You evaluate them, take the good ones and add them to the baseline. When it is at a point that you would say "yes ... now 80% of them look like stations".
You have your station algorithm with your parameters. You take a minimizer that iterates through the possibility space of your parameter set and uses the neural network as cost function to evaluate the outcome of each parameter set. This way you could narrow down the local or global minima minima for each new iteration of the ship algorithm over night.
...
I just realized hyperGAN does not work this way, does it? It creates images, it does not recognise them? Well then .... there sure is some free code for learned image recognition out there. ^^
(hypergan could work by mapping the parameters of good looking stations on pixels of an image (with brightness representing the parameter value) and than using hypergan to spit out new parameter sets)
...
Obviously I have no idea what I am talking about XD
It's way to late here. I should sleep. Good night. ^^
Ah, oh, wait!
Lindsey, love your devlog! ( like every LT devlog ... don't feel too special
)
But yeah, what Victor said. Waiting for something awesome is a trap. ^^
Machine learning and neural networks are not the solution for Limit Theory.
Y'alls ideas are interesting in theory, but they require a lot more time to implement than it may seem at first glance. Let's break down what these solutions involve:
1) Writing algorithms to generate ships & stations (which is what I'm currently doing)
2) Writing those algorithms work in the context of a neural network that can change the algorithms
2) Evaluating the output of those algorithms, either:
- By writing an algorithm to evaluate the output, and/or
- By building a tool to allow people to assess the output by hand
Doing 1 without doing 2 is already enough of a challenge that it will take the time that we have to finish Limit Theory. Step 3 has the added difficulty of being subjective- whether or not stations "look good" is up to the viewer. The vast majority of problems that people solve with machine learning have objective, discrete solutions: the robot learning to walk either remains upright or not. The Google image recognizer either guessed the image correct or not. Those problems have a clear win/fail state, and when that state needs to be evaluated by hand, the company (like Google) often has hundreds of thousands, if not millions, of people happy to give many millions of pieces of data of evaluation of the solutions.
Every possible tool and solution one could use when building something comes down to a cost/ benefit analysis. Here, the cost is time, and the benefit is time and quality. I'm confident that both: 1, my current method of approach and the tools we have will continue to produce awesome, quality, looking stuff; and 2, the time cost of implementing machine learning would use more time than it would save.
Although machine learning won't be a part of building LT, who's to say that machine learning can't do all of this stuff in the future, given unlimited time and money?
Y'all aren't wrong that it's A solution- it's just not the solution for us.
Ship Inspiration Pinterest!! (send me stuff)
"You’ve got to work on something dangerous. You have to work on something that makes you uncertain. Something that makes you doubt yourself... because it stimulates you to do things you haven’t done before. The whole thing is if you know where you’re going, you’ve gone, as the poet says. And that’s death."
- Stephen Sondheim