1st Render Code Images!!!

Officially getting images of my data set out of the render code!!  Next steps are to explore the look dev options already within the code and modify to get what I need.

radius = 0.5

different frame, radius = 0.1

my first image :-)... its so beautiful *tear*

Getting the Render Code to work...

In order to render the millions of particles I'm going to need efficiently, I'm going to be working with some C code created by Frank Summers at STScI.  He uses it to render stars, gas, etc for press release visualizations, and has recently reworked it to take some of the work we did this summer into account.  I'll be modifying the code as necessary for my purposes.

Step 1 of getting my particles into the render code, is to get the render code to work on my machine.  After a couple hours of monkeying around and some help from the VIZ help desk, the code was able to be run successfully.  Here's the test image I got out:

According to Frank, this is exactly what I should be getting.  YEAH!

*** Update: 2nd test with both gas and star splats.  1920x1080 resolution, ~ 13.5 minute render time.

Look Development, Set 3

Finally culled out sets that are basically correct.  The low mesh has an extra subdivision added in by accident, but overall, all look parameters are operating as expected!!! Onto the render code!!!!

*** UPDATE: fixed the issue with the particles that were out of place.  Everything's looking good.  Will still need more look tweaks as we go... I may have clamped the alpha too much on the top layer.  Also, the camera has a lot of flexibility, but won't be able to get parallel with the mesh- that's when the illusion breaks and the opaque particle layer starts to create harsh edges.  Anyway, here are the pretty images!

both, from above

both, perspective

both, close-up... pretty :-D!

low, from above

==============================================

both sets operating as expected (minus the extra low subset)

both, perspective

top mesh, above

top mesh, perspective

low mesh, above

low mesh, perspective

Look Development, Set 3, Again

Top looking pretty good.  Bottom has issues, need to recull.

Just the top.  Going to add a little falloff for the opaque underlayer so it doesn't look weird.

Close up on the top layer.

Look Dev Redux!

In an effort to fix the mixing of the 2 sets of particles I was getting previously, I've made some changes to how I'm generating particles and creating transparency. 

For the top mesh, I'm going to create a set of particles from -0.3 -> 0.0 below the surface that are completely opaque.  This set will only exist where the actual cloud is opaque.  The next set of particles will exist only above the surface, with ramped transparency as they get farther away (just like before).

For the lower mesh, I'm going to move the mesh downwards a bit, so that the particles don't exist in the same area.  Then, I'm going to only use particles that are below the surface.  The opaque particles will exist from -1.0 -> -0.7, with ramped transparent particles above. 

By using this method, I'll be able to have opaque-ness where its needed, and still be able to play with the falloff/transparency to create a wispy cloud appearance.

Some tests below:

Top Mesh- opaque particles are underneath to the right.

Checking opacity.

Particle Sets So Far

I mapped the top views of the different sets to the original image to see how much coverage I had so far.  Also noticed that the alpha for the top mesh is spastic... too harsh in some areas, fine in others.  Going to take a few tweaks to get it right.

original

with particle subsets

Look Development: Set 3

Here's some images from subset 3.  On their own, each set of particles looks pretty good.  Together, we have some issues with them blending together.  The lower particles are visible in the upper ones, creating blue streaks.  I tried to make both meshes thin, but that didn't fix the problem.  Maybe a thickness map of some kind?  To determine what PDF to use?  I also feel like the nebula overall is too thick, so I may switch to a more restrictive PDF- (x-1)^10. 

Thickness: 0.2

Thickness: 0.2

Thickness: 0.9

Thickness: 0.9

Top only- note the crappy alpha...

Top only

Bottom only- note the bottom corner cut out...

Bottom only

Look Development: s106 Set 2 (cont'd)

Removed the middle layer of particles.  Everything is looking pretty good.  Moving on to a final (larger) subset before trying out the render code.

Look Development: Set 2, cont'd

In an effort to isolate the black particles of the 2nd mesh and keep them from muddying the other colors, I used a quadric PDF on that data set.  By doing so, I created an inner core of darker particles that didn't cause too much damage to the outer mesh.  That said,  I inquired about the purpose of the dark mesh, and it appears I can eliminate it, as it was a helper mesh for the comp.  Anyway, here's where I got with it before that.

No transparency... still speckly.

Size variation and transparency gets rid of speckles.

Inner core of black particles.

Close up of inner core.

Look Development: S106 Set 2

Here are my tests with the second subset of s106.  The dark particles from the 3rd mesh are intermixed with the brighter ones, which is causing issues.  I'm going to try using a more severe PDF (quadric) and see if that tones them down a little.  Otherwise, I may need to discuss keeping them.  They just muddy up the color values right now.  Below is the combined set, followed by the top, middle, and bottom meshes.

No transparency- you can see the dark speckles.

All 3 layers :-P.

Top

Middle (source of the dark speckles)

Bottom

Look Development: S106, Set 2

I have now added in the 3rd mesh of s106 and I'm using a separate set of particles to test out the look parameters I've developed.  I discounted the failed particles in culling these sets out (simply removed particles that were completely transparent).  I tested a single subset to see approximately how many particles I'm losing to this error.  I then extrapolated how those percentages would affect the overall set.

Subset test:

w/ fails removed
101900
w/o fails removed            #OK             #FAIL            %
103883                                 102233       1650              1.59%

Extrapolation for entire set :

w/ fails removed
1900063
w/o fails removed             #OK             #FAIL            %
1930274                                1900063     30211            1.59%

If you take into account that we're applying a PDF to our data set, some of these lost particles would be removed anyway.

Original Set: 1900063
PDF Set:         632123 (30% retention)

Viable Fails:  9063

Not sure how big an effect 9K extra particles would have on the appearance of the data set.  However, the ideal situation would be a solution to the texture-lookup fails.

S106 Structure

S106 is made up of 3 main meshes with different texture maps.  I've been working with 2 so far, and I need to start incorporating the last one.  Unfortunately, 2 of the meshes have the same geometry, just different texture maps, so the culling code won't necessarily choose the correct mesh for the texture lookup.

I've decided to break up the particle generation and culling into separate chunks, one for each mesh.  I'm also going to add another custom attribute to define which set each particle is from (1, 2, or 3).  This way, I can tweak the look of each set separately in the render code.  After culling, I'll combine the PDA files with my combinePDA code.  Hopefully, this will also make the culling code more efficient since the particles will be generated only for a specific mesh.

Here are the different pieces to the S106 nebula:

TOP

MID

BOTTOM

COMBINED

Comparison: Original Texture File and Particles

Just to see where we're losing detail and if its an acceptable loss...

Look Development: Current State

Starting to look pretty good... Next step is playing with the noise parameters to see if the color can be broken up a little more.  Also, I'm missing a layer of particles with this- the "dark" particles.  So I need to add that in too.

Playing with the noise to see what I could get.  I'll have to investigate how to calculate different types of noise using C instead of Maya.  Here are the attribute changes I made in Maya though:

Roundness = .288

Noise Freq = .259

I may still need to adjust the roundness attribute back up so that I don't get the patchy dots. 

w/ default noise attributes

w/ changed noise attributes