Photon Room

[ypoissant]

Here is a work in progress. This a kid room. The project is a collaboration with Tony Matias (who signs Tony on the list). I plan to use this scene as a basis for the "Interior Global Illumination using Photon Mapping" tutorial.

 

This is not quite the latest render but close to it.

 

Any comments are appreciated.

[JTalbotski]

That is a gorgeous render! Fantastic detail. I love the lighting. There's so much depth in this scene.

 

 

If I had to find something to change, the shoes look pretty big for someone who sleeps in that bed.

 

Jim

[Obnomauk]

Yowsers very nice render

 

overall it's fantastic, I have a couple nits about the content of the room but none for the execution.

 

Primarily I find that some of the room looks like it belongs to a kid and some of it looks like an adult lives there. the curtains in particular seem awful adult... but again that's just a nit.

 

I'm not sure what's going on with that chrome robot there but it is a nice focus to the piece.

 

Great work!

-David

[MixePix]

Very nice! How long did it take to render?

[Zaryin]

Wow, I can't wait for the tutorial on this. The lighting is exceptional.

[modernhorse]

My only *problem* with this piece is the content of the room as well. It jumped out at me as slightly odd. The curtains, rug and bedspread look more girly or adult. The toys and sneakers look like a boy's. Then on the door it looks a like a poster that might be in a girls room.

 

But all that being said ... it is quite stunning and beautiful.

[ypoissant]

Thanks for the comments everyone.

 

Jim - Yes, the shoes are a tad big. I've since resized them and added some laces. Next render.

 

David - There is indeed a culture clash in this room. This is on purpose to add tension. There is the adult culture that can be seen primarily in the decoration (the wallpaper, the curtains, the frames) and in some tidyness. And then there is the kid's culture who tries to appropriate the space (drawings on the wall, toys on the floor) and the general untidyness. The poster is a concession from the adults. And yes, the robot is supposed to be the focus of the image. I tried my best, with composition elements (bed, bat, bus, shoes, door frame) to direct the view to it. But I don't quite know what the robot is after .

 

MixePix - I'm not sure you want to know . It took 31 hours on a P-IV, 2.6mHz, 512 meg. The render was 1200 x 676 (I resized it in Photoshop for posting. It should be about 8 hours at the intended size of 600 x 338. I used the highest Photon mapping settings mainly because of the horizontal blinds. Without the blinds, I could probably reduce the final gathering samples from 500 to 150 which would drop the render time even further to about 2 hours.

 

modernhorse - The inhabitant of the room is a girl. But she likes to play with the boys around and is enrolled in boys games such as baseball. But I agree, there should be more girlish props in this room. Any suggestions for props?

[ypoissant]

I thought I'd also post these renders for comparison.

 

This one of with the ceiling light. (5 bulb lights). This one is a very early render.

 

This one is with additional lights inside and around the room to somewhat simulate indirect illumination. There are 16 additional lights positionned close to the walls.

[MixePix]

Ouch! I always forget some small thing and end up re-rendering the whole thing. With that kind of rendertime I would have killed myself long time ago. 2 hours sounds a lot better though.

[Iham Wrong]

Simply awesome. The shoes are the centerpiece BTW. Next the bed, rug and finally the hulahoop. The turtle sticks out a tad which leads to the door dimensions that seem a tad thick. Then back to the nightstands and lamps with seem a tad thin. Even with scale issues the room is excellent, but I'm in no position to critic.

The lighting is stunning and as near photorealistic as seen anywhere, but I can sense that the venetian blinds are driving you batty since they won't render correctly.

 

Maybe not the nightstands. The lightswitch by the door is missing.

[jfirestine]

Very impressive render. Can you post some wireframe images?

 

I wish my kid's room was that clean!

 

 

Very Nice!!!!

[xor]

Now that's some good lighting. Just what I'm looking for. I think many of us would be interested in those.

 

Keep us posted on your tutorial.

[Mega]

Let me throw my $.02 since the room is somewhat, loosely, based on my daughter’s room. By the way…it’s never that tidy!! Yes, that is the size of the bed that my wife and I bought her...fine Italian wood! I modeled the bed, and the rest of the furniture, to scale. My daughter likes to be "free" while she sleeps. Also, while my daughter takes dance and plays with Barbies...she also plays a wicked game of tennis, baseball, soccer and loves Harry Potter on the X-Box. She has a fondness for anime due to me!! There is much tension, in a fun and loving way, between adult and child because, let's face it, kids are tough nut to crack! Especially a five year old. My daughter drives us batty because of that reason. My wife can’t believe that she sat through Star Wars with me…go figure. The room, and the chrome robot, are for an animation that I am working on. I am currently modeling the girl, several other characters as well as adding more specific models to make the room a wee bit more…girly.

 

When Yves and I spoke about photon mapping, we decided to use the "room" as jumping point - since I had most of it modeled anyway. For me, it was a no brainer...I jumped at the opportunity to work with him. Yves, besides being an incredible artist, is incredibly knowledgeable and a complete professional as well as being completely supportive and enthusiastic. I have learned quite a bit from him...

 

Yves really turned the room around, greatly improving it, from the original scene that we begun with. He really made the room click and added much mood and detail to it.

 

I hope I shed some more light on the subject. If not, please ask away…

 

Tony

[robcat2075]

I think it's a fabulous demonstration, appropriate decor or not.

 

Based on your notes for the other images, the top one is the result of one light source for the whole room, while the others used more lights to simulate the diffuse lighting effect?

 

Is "photon mapping" another term for "radiosity". A part of it? Something else entirely? I ask because...

 

If Radiosity is On, photon maps are automatically created ...

 

...is the only mention i found in the docs.

[ypoissant]

You are correct. The first image is the result of one ceiling light furniture which is actually 5 lights arranged as 5 point a candelier. The darker image is exactly the same lighting with no additional lights to simulate GI and the other image is with additional lights to try simulate GI somewhat. I could certainly tweak the additional lights to get a more radiosity like lighting. In particular, I courd use multi-ray soft shadows but I wouldn't be able to get the nice warm colors I get with photon mapping without tweaking from the start the color of the different surfaces.

 

Photon mapping is a technique for producing global illumination. Radiosity is another technique. But since radiosity was the first technique ever devised to produce global illumination (actually radiosity could only compute indirect diffuse illumination), "radiosity" is now often used as a generic term to signify GI. In other words, "Radiosity" is now used to mean what the first researchers had in mind when they developped the technique, which is to accurately compute all lighting effects, including direct and indirect illumination, from fully diffuse to fully reflective light bounces, reflective and refractive caustics, participating media, subsurface scatering and more. Of course, the actual radiosity algorithm results where very far from their dreams.

 

If you want more informations about photon mapping in A:M, please visit my Photon Mapping Tutorial page.

[JackMcRip]

Fantastic !

This shows the capacity of A:M !

Thanks for this great Work.

[robcat2075]

It took 31 hours

 

That reminds me of the render times I used to get with Sculpt 3D on my AMIGA.

[kanga]

Hi Yves!

 

How ya doin? What brilliant warmth and atmosphere you generated here in this render. Right now I could use a bit of sleep myself,.... an if I could,..... just cuddle up next to that bear,......zzzzzzzzzzzzzzzz

 

The Force be with ya,

[JohnArtbox]

Beautiful Yves & Tony,

Out of curiousity how long did the 16 light version take to render? And can you bake the photon lighting?

[macb]

hehe...

I have to add a "me too post" to THis Amazing render!!!

 

Excellent job Yves

 

[Mendi]

Congratulations, it's a wonderful render. At first sight I wondered if it was an image reference or a render.

31 hours is not so much time, seeing what the results are .

[robcat2075]

a small issue... but why does the reflection in the floor of the legs of the chair and the bed look so much brighter than the "real" legs?

[ypoissant]

JohnArtbox - OK, I had to rerender in order to get the render time cause I didn't note it before. So for a 1200x676 antialiased, the 16 light GI faker took 1h15m. I forgot to mention that the photon mapping render setup was with multipass set to 16 too. I will launch a 16 multipass render of the faker tonight and report the render time later. And also, I plan to do a render with 10 multi-ray cast soft shadows. This should give results somewhat nearer to GI but will probably take about 10 times longer. So given those numbers, photon mapping is not that long after all.

 

No. It is not possible to bake the photon lighting. This was something possible with the actual "radiosity" technique because it actually produced a color map of all the surfaces in the scene. But although the name "photon mapping" contains the word "map", this map is actually a 3D data structure which holds the photon locations in the scene along with their color and intensity. The closest it can get to baking the information is the "precompute irradiance" step.

 

robcat - The reflections appear lighter because of the floor color. Currently, in order to get normal reflections, the reflective surface color needs to be black. I had the same issue with the mirror which is 100% reflective but because its color was 50% gray, the reflected image was lighter than the actual image. Until I set the mirror color to black.

[zandoriastudios]

Great image!

 

Is the first image only global illumination--using the 5 lights on the ceiling?

The 3rd image you said had 16 additional lights to fake global illumination --is that for a comparison with the first image, or does the first image also have the extra lights?

 

I was just confused about the number of lights in the first picture.

[ypoissant]

Image 1

• 1200 x 676 pixels
  
• multipass at 16 passes
  
• 5 lights arranged as a 5 point candelier located at the center of the ceiling about 1 foot from the ceiling. Intensity set at 50% and the falloff set so that it does not touch any object in the scene. Color lightly yellow. Shadow ON ray-traced with only 1 ray and shadow darkness set to 100%
  
• Radiosity ON, 1 000 000 photons, 15 bounces,
  
• sampling area: 1600 and 500 photon samples,
  
• Final gathering ON with 500 samples and 50% jitter
  
• Precompute irradiance ON
  
• All surfaces have ambiance set to 0 and radiance set to 100% (although it would be more realistic to set radiance to something around 95%)
  
• Render time: 31h
  

Image 2

• 600 x 338 pixels
  
• No multipass, it uses the default antialiasing
  
• The same 5 lights candelier as for image 1 except intensity set to 100%
  
• Radiosity OFF
  
• Same surface attributes as for Image 1
  
• Render time: I didn't note
  

Image 3

• 1200 x 676 pixels
  
• No multipass, it uses the default antialiasing
  
• The same 5 lights candelier as for image 1. Intensity set to 50%
  
• 15 additional orange lights, 7% intensity and 1500cm falloff so that there is no light intensity attenuation for the entire space of the room. The lights are arranged as follow:
  • 4 lights in the room corners near the ceiling
    
  • 4 lights at the center of each wall near the ceiling
    
  • 4 lights in the room corners near the floor
    
  • 3 lights at the center of each wall near the floor (I didn't put a light behind the bed)
    

  [*]Radiosity OFF

  [*]Same surface attributes as for Image 1

  [*]Render time: 1h15m

Image 4

• 1200 x 676 pixels
  
• Multipass set to 16 passes
  
• The same 5 lights candelier as for image 1. Intensity set to 50%
  
• 15 additional orange lights, 8% intensity and 1500cm falloff so that there is no light intensity attenuation for the entire space of the room. The lights are arranged as for image 3
  
• Radiosity OFF
  
• Same surface attributes as for Image 1
  
• Render time: 8h30m (This kind of put the 31h into perspective.
  

I don't think I forgot anything

[HeadlessBill]

That first image you posted is amazing. I've been trying to get renders like that for the office set for the short I'm working on, but nothing anywhere near the results you are getting. I'm looking forward to your tutorial when it is done.

 

31 hours for the render. Seems to make it unusable for animation (but I'm gonna try, dammit!)

 

Oh, the questions I would have are:

 

1. I'm assuming you are using bulb lights, correct?

2. Is this done using the 10.5o release or the Alpha 11 release?

[ypoissant]

Here is another image rendered last night.

 

Let's call it Image 4. I've edited the list of image caracteristics 2 post up to add a description of this image. This is not rendered with Photon Mapping. I upped the environment lights intensity so that the bed and the chair does not look so dark but the walls are now washed out.

 

BTW, I also plan to do at least 3 more photon renders:

1) with the table lamps ON and the ceiling light OFF.

2) with the night lamp ON and the ceiling light OFF. There is a night lamp between the bed and the table desk

3) with the closet door open, the closet light ON and the ceiling light OFF.

 

All lights are bulb lights and all images are rendered in v10.5.

[shaunf]

Supurb Yves. Particularly the first image as I almost feel I can walk into the room and put on those shoes....

(I have big feet)

[Zaryin]

31 hours is a long time to wait, but you sure do get a better image, in my opinion. I can't wait ot see the next photon rendering.

[chazcron]

While glorious, the lighting lacks an identifiable source. I'd like to see that bedside lamp illuminated.

[HomeSlice]

Yves, you're an absolute genius. You've been very busy lately. Your final render is eye popping. You have a knack for approaching these things in a methodical way so the rest of us can learn from your experience. All I can say is WoW, and Thank You for the information.

 

Holmes

[kualrobi]

Impressive! excellent image,

 

30 hours times 30 frames times 30 seconds or more, god I wish I had a renderfarm.

 

Got a question about the pictures framed on the wall. Are the photos themselves so dark, or is that the result of the GI and glass? Sorry, saw the answer in the later render. Tried a skylight dome on eyes once and the iris and pupils became a big black dot.

 

What I think is great too is the double reflection on the two glass panels of the windows, [the front one a tad tilted] so is the wood on the bed. Perhaps there is a bit much perspective in the top part of the door though <_ but hey that details....perhaps some crap shoved under the bed adds to untidyness appearing be tidy..>

 

Chapeau!

 

Arthur Scheijde

[win XP / 2.1gig / 512Ram /v9.5]

 

ps: Do I see the turtle character of the new PDI animfilm Madagascar on the table in the corner in the middle, or is that the far out "I just waxed it.."surfer turtle from Pixar?

[ypoissant]

Arthur - I also noticed the darkness of the framed pictures as well as their cutout. It is probably due to the interaction between the photons and the attribute of the glass (refraction index and color) I will have to adjust those attributes. Setting the glass color to black there might not be such a good idea in this particular instance.

 

The shifted reflection is simply the result of modeling a sliding frame window.

 

Yes, Tony also suggested to add some objects under the bed. I will see what I can do with that.

 

I also forgot to mention that the bed, the chair, the windows, the door frame and the moldings wood textures are procedurals but the door itself and the floor are maps. The wallpaper is aslo procedural.

 

I don't know where the turtle come from. Tony?

[Mega]

I hadn't seen any new film from Pixar that featured a turtle. Where can I see this? The turtle, if I remember correctly, was either from the Hash CD, or from a free site generated by a generous AM'er!

 

As far as clutter, specific to under the bed, when I get a bit more time on my hands, I plan to finish some models that I hope to populate the room. I'm looking at my daughter's room and trying to decide which objects to use in a clever manner.

 

The great thing with working with Yves is that he really pushes the artist, that being me, to do the best that I can!!!! Believe me, the room is far better than I could have hoped for. There are several things that I hope to add to the room...I only wish I had more time.

 

There are still many things that I hope to do in prepping for my animation, but Yves and the photon mapping technique have helped me lay the stage for this. I encourage all AM'ers to really try, and push, photon mapping. This way we can really make A:M shine and allow photon mapping to be better developed for future releases.

[KenH]

That's a beautiful room! Great work! It's also encouraging to hear that your AM works non-stop for a day and a half without problems.

Given that I'm doing a simple room at the moment, I'd be interested to know if you used "light lists" in this image. ( I didn't see it mentioned in the posts I browsed)

Cheers.

[ypoissant]

I've had A:M v10.5 open non-stop, either editing or rendering or idle, for about a full week without any kind of problems. Even with Photoshop, Outlook express and other utilities opened at the same time.

 

There are no light lists in this scene.

[cronos]

thank you for this incredible Photon mapping tutorial !!!

 

my first tests , with many errors

[Sevenar]

Yves & Tony: Simply beautiful work. This is the kind of image that belongs in Hash's new marketing strategies...

[robcat2075]

I've had A:M v10.5 open non-stop, either editing or rendering or idle, for about a full week without any kind of problems. Even with Photoshop, Outlook express and other utilities opened at the same time.

 

So what is this magical alignment of system components?

 

OS, motherboard, RAM, RAM type, CPU, videocard... all of that. Are there things you've learned not to install?

[ypoissant]

My current system is brand new. ASUS motherboard P-IV 2.6mHg processor, 512meg Kingston RAM and ATI AIW Radeon graphic card (Which shows its age and seriously need to be replaced). Softwarewise, Win XP Pro and I didn't limit applications installation either except for Visual Studio. I decided that there would be no software development of any kind installed on this computer.

 

My previous system was a Dell-PIII 800mHz, 256meg 3 year old with the original Win98SE. The graphics card was replaced by a GeForce-II. And a ton of software of all kind. Even though it became really unstable with time, I didn't have more stability issues with A:M than with any other applications. Actually, Corel and Macromedia products where way more unstable on this computer. And Visual Studio was the application which crashed my computer the most often (but that is to be expected when you think about it).

 

Sorry, I can't be of more help there.

[OdinsEye2k]

A beautiful render, indeed. You say that the GI helps to warm up the room - does that mean that GI now includes color with the photon bounces?

 

In other words, the lighting from GI should now automatically match the 'temperature' of the room depending on the dominant color, yes? I.e., an orange room would have 'warm' lighting and a blue-grey room would be 'cool.'

[zandoriastudios]

look at the cornell box image posted by Cronos... You can plainly see the relflected color

[ypoissant]

The Photon Mapping algorithm takes the color of the surface from which it bounces into account. Not only the hue of the color but also its value and saturation. You further can control how much each object absorbs or bounces photons with the Radiance attribute and you can control how many maximum times you want your photons to bounce.

[ypoissant]

Almost one year ago, I promised, in this thread, to post more renders of this room with different light settings. Here they are. All done with v11.0t :

Photons cast : 1 000 000

Sample Area : 1 000

Photon Samples : 200

Max Bounces : 5 (I need to try more)

Final Gathering Samples : 500

Jitter : 100%

 

Average render times at 25 passes on a 2.5gHz WinXP : 6h30m

 

Only one table light

 

The two side table lights

 

Only the tiny night lamp behind the desk

[Sharky]

Hi!

 

Congratulation!

It is very beautiful work and very detailed!

 

Cheers...

 

Sharky

[D.Joseph Design]

WOW! I'd love to do that with my room scene, but render times are already way too high at 1.5 hours per frame.

[Dearmad]

Beautiful lighting in the second image- really nice.

 

If you "need to try" more bounces, how accurately is attenuation handled by the mapper? Is it manipulable beyond the radiance attrribute? Is it right to interpret that you need to set radiance for each surface/texture in order to manage that? Or is it for an entire object? If it's attached only on a per object basis, all sorts of strange (unreal) results could follow.

 

Seems like with >5 bounces or so and dimmer lights, you run into perception vs. physics issues- the human interpretation of sensation differs from the "camera", no?

 

Lots of interesting stuff though.

 

Interior designer crit: With the lighting so beautifully done, that grey cabinet sure needs to go down into the garage, really sticks out like a sore thumb with such nice lighting... Needs a paint job!

[dingo]

Average render times at 25 passes on a 2.5gHz WinXP : 6h30m

Hey Yves,

 

This may be a dumb question but:

 

How did you determine to use 25 render passes? Instead of 10 or 100? Why was 25 the sweetspot for these renders? Or was it?

 

Peace,

dingo

[ypoissant]

Dearmad,

 

How accurately is attenuation handled by the mapper?

 

As accurately as a photon can be. Or, in other words, not attenuated at all.

 

The detailed explanation is this: Photons don't loose energy They are either absorbed or bounced. In ray-tracing, we apply the inverse square law of attenuation because we are not dealing with photons but rather with the photon density. So as light travel a distance from an emission point, the photons are more and more dispersed and thus they are less dense. That is the inverse square law. But with the photon mapper, we don't compute the density, we estimate it. We shoot a million photons in a scene and then estimate their densities everywhere in the scene. The end result is as if you had positionned a million light sources in the scene.

 

In real world, photons have different level of energy which corresponds roughly to the color along the spectrum. In CG, each photon is actually 3 photons. R, G and B photons represented in one single CG photon. And their absorption vs bounce is computed on the photon by modifying the RGB energies. There is a part which is done statisticaly and a part which is done computationally. When a photon hits a surface, the decision to bounce it or absorb it is done statistically. Once this decision is made, the photon RGB energies are adjusted according to the surface diffuse color.

 

Is it manipulable beyond the radiance attrribute?

 

The parameter that have the most influence is the surface diffuse color. That is what adjusts the photon RGB energies and that is what determines the statistics for deciding if a photon is bounced or absorbed. Other properties such as reflectivity, transparency, etc have some effects mainly on the way or the pattern the photons are bounced.

 

Is it right to interpret that you need to set radiance for each surface/texture in order to manage that?

 

The radiance should actually be set to 95% on all surfaces if we wanted to be more realistic. That is because, in reality, even a very white surface have a reflectivity of 95%. But 100% is very good for CG purposes and you wouldn't notice the difference anyway. Think of the radiance as a constant factor that may be used to attenuate the RGB characteristics of the surface. You would normally not use that and just leave them at 100%. The actual reflectivity of a surface is computed from its RGB diffuse color.

 

Or is it for an entire object? If it's attached only on a per object basis, all sorts of strange (unreal) results could follow.

 

Indeed, if you adjust radiance differently on a per object or per group basis, you will likely get all sort of unreal effects. This could be an artistic exploration and if mastered could probably lead to some interesting artistic style. But, personally, given the time required to test a photon mapping scene, I would not attempt that.

 

Seems like with >5 bounces or so and dimmer lights, you run into perception vs. physics issues- the human interpretation of sensation differs from the "camera", no?

 

The greater the number of bounces, the more interreflection between objects in the scene. At 0 bounces, the scene would be illuminated equivalent to standard ray-trace with one light. At 1 bounce, there would be some interreflection but still the shadows would be quite harsh. At 30 bounces, you start to get an illumination that looks like reality. In reality, photons can bounce several millions of times.

 

With the lighting so beautifully done, that grey cabinet sure needs to go down into the garage, really sticks out like a sore thumb with such nice lighting... Needs a paint job!

 

The cabinet is actually purplish like the two other cabinets. The grayish look comes from the specular reflectivity from the spot of light.

 

Dingo,

 

How did you determine to use 25 render passes?

 

25 passes is determined purely on the antialiasing that is necessary for the blinds blades. I would use 25 passes even wiithout photon mapping in this scene because of the blinds. And I determined it by trial and error. The number of passes does not affect the photon mapping quality since the final gathering samples are distributed among passes. In other words, if you set 500 final gathering samples and 25 passes, then each pass will average 20 samples.

[ypoissant]

Here are some Photon Mapping night lighting. Render time aroung 5h15m each.

 

One klieg light for moon, no softness. Angle tightly fitting the window frame. Than another klieg light, same position and same angle as moon light but pointing exact oposite direction at a 100% reflective patch.

 

The light pointing at a mirror and then back into the room is to allow photons to be stored at first hit. Normally, the first hit is nor stored because the first hit contribution is computed with Ray-Tracing but when light is outside the room, the scene is better illuminated if first hit is stored.

 

Twice the same light setup. One for each window.

 

The only light in this scene is the night light behind the side desk.