Title: rise Post by: lycium on September 13, 2007, 05:23:26 AM http://lyc.deviantart.com/art/rise-64751071
a cropped+resized preview (see link above for the full 1920x1200 image): (http://www.fractographer.com/wip/allura_fractal.jpg) Title: Re: rise Post by: Nahee_Enterprises on September 15, 2007, 07:11:47 AM Very Nice!! I like your Sierpinski Sponge very much. :)
I understand the blurring used for the area furthest away from the viewing point, but I question the blurring used for the area closest to the viewer (unless, of course, you were trying to simulate a camera lens that has focal point towards the center of the object). Anyway, great work!! Title: Re: rise Post by: lycium on September 15, 2007, 08:32:31 AM thanks :)
I understand the blurring used for the area furthest away from the viewing point, but I question the blurring used for the area closest to the viewer (unless, of course, you were trying to simulate a camera lens that has focal point towards the center of the object). i'm doing exactly that ;) btw, since this appeared in my deviation "23", it's been used for all the new 3d apophysis fractals... coincidence? you be the judge: http://comments.deviantart.com/1/62663760/521757431 besides the light bouncing around inside the fractal (potentially infinitely!) there's a thin-lens camera model, so i set the focal depth, aperture size, field of view, ... just like a normal camera. since this is a spectral renderer i could eventually integrate "real" cameras, via measurements like these: http://www.graphics.cornell.edu/online/measurements/sensors/index.html i did a fun little test with the depth of field later (http://lyc.deviantart.com/art/allura-text-64901142) and it's also used in my "snow" deviation (http://lyc.deviantart.com/art/snow-64216165). Title: Re: rise Post by: David Makin on September 16, 2007, 12:58:01 AM Excellent work Thomas !
I'm lazy sometimes - is there any chance you could let me know the correct coords. of the vertices ? (I'd like to try this using my 3D escape-time IFS formula in UF) Title: Re: rise Post by: lycium on September 16, 2007, 01:11:40 AM Code: static float icosaSierp::baseVertices[] = i'm still really interested to know how your UF algorithm works btw ;) Title: Re: rise Post by: David Makin on September 16, 2007, 01:55:25 AM I can send you the formula file if you like - it's just a text file.
Title: Re: rise Post by: lycium on September 16, 2007, 02:06:26 AM hmm, actually i should be studying for my exams :( thanks for the offer, but i should exercise a little restraint -- my last two works were already done in "borrowed time" (i find it's much easier to be a productive programmer if i'm supposed to be doing something else!).
i'll have to look into it later, say novemberish :) Title: Re: rise Post by: David Makin on September 16, 2007, 02:22:23 AM OK !
Title: Re: rise Post by: David Makin on September 16, 2007, 03:30:00 AM Just tried a Sierpinski pyramid rotation:
http://www.youtube.com/watch?v=moCJUlQf0Fo Rendered 1 rotation in 101 frames at 320*240, took 1hr 40mins Sorry only diffuse camera lighting and no shadows at the moment. Title: Re: rise Post by: lycium on September 16, 2007, 05:30:58 AM pretty interesting colouring, i suppose it's done based on the traversal steps?
btw, my lighting is also purely diffuse, but all possible light bounces are accounted for (unbiased estimation of infinite recursion) which is what illuminates the deep interior regions -- and what is responsible for the extreme render time :| Title: Re: rise Post by: David Makin on September 16, 2007, 01:55:16 PM In this case the colouring is based on the distances travelled on just the two final steps, the formula allows you to use more if you wish and also allows colouring based on the positions rather than the distances. I'm also hoping to add colouring based simply on the genetics.
Of course the 1hr 40mins was for 2 layers, 1 lighting and 1 colouring - it's possible to write it so this could be done in a single layer using UF's direct colouring option but I haven't got around to that yet :) Thanks for the coords BTW, actually I had a "duh !" moment and looked up "icosahedron" in Google and found the coords on WiKi. Title: Re: rise Post by: David Makin on September 16, 2007, 06:59:54 PM Tried it using my escape-time IFS formula:
http://makinmagic.deviantart.com/art/Sierpinski-Icosahedron-65047470 Title: Re: rise Post by: David Makin on September 16, 2007, 07:17:57 PM BTW there's a cheap trick you can use to approximate (very roughly) the effect you achieve by bouncing the light around and you can do it with just the normal first pass - instead of the light value going from 1 (full-lit) to -1 (completely the wrong way) change it so the scale goes from 1 (full-lit) to 0 (completely facing the wrong way) and reduce the ambient level somewhat - the result won't be as good as your image here but it will reduce the render time considerably :)
Title: Re: rise Post by: lycium on September 16, 2007, 11:14:13 PM In this case the colouring is based on the distances travelled on just the two final steps, the formula allows you to use more if you wish and also allows colouring based on the positions rather than the distances. I'm also hoping to add colouring based simply on the genetics. i'm not sure what you mean by the genetics, but colouring based on the last steps opens sufficiently many options (angles, velocity, distances, ...). Thanks for the coords BTW, actually I had a "duh !" moment and looked up "icosahedron" in Google and found the coords on WiKi. np, and you at least stated upfront that it was out of laziness ;) Title: Re: rise Post by: lycium on September 16, 2007, 11:23:45 PM pretty cool how they look so similar! it's possible that i rendered mine with way too much recursion depth (10 or 12 iirc), which could also be responsible for the numeric precision issues i was facing...
BTW there's a cheap trick you can use to approximate (very roughly) the effect you achieve by bouncing the light around and you can do it with just the normal first pass - instead of the light value going from 1 (full-lit) to -1 (completely the wrong way) change it so the scale goes from 1 (full-lit) to 0 (completely facing the wrong way) and reduce the ambient level somewhat - the result won't be as good as your image here but it will reduce the render time considerably :) ah yes, that's a standard trick used in game development :) a reference that immediately springs to mind is the mighty charles bloom, superstar game developer and theoretical physicist turned poker player, in his 3d rambles here: http://cbloom.com/3d/rambles.html (3-13-03 entry). Title: Re: rise Post by: David Makin on September 17, 2007, 12:27:41 AM [i'm not sure what you mean by the genetics, but colouring based on the last steps opens sufficiently many options (angles, velocity, distances, ...). My genetics algorithm siimply assigns a value to each transform and combines these numbers based on the genetic history (i.e. the most recent transforms) for each point, it's more rigid in form than using the locations or positions or angles etc. but can be useful - e.g. if only the final transform is considered for each point then your colouring seperates the object into the final genetic areas/volumes. Title: Re: rise Post by: lycium on September 17, 2007, 12:31:20 AM i see. that's pretty much how the flam3-colouring algorithm works, and i notice now that it shares the "genetics" term... how did i miss that? :P
Title: Re: rise Post by: David Makin on September 17, 2007, 03:53:50 AM Also tried an anim of a 4-sided Sierpinski Pyramid:
http://www.youtube.com/watch?v=1m5rFdpI3GM Title: Re: rise Post by: lycium on September 17, 2007, 04:16:50 AM looks interesting, youtube is pretty brutal with the compression though :/ |