Cloudmap ????

[cartrite]

This is true and a split up full sphere mesh with no 3d details would be useless if Chris implimented normalmaps for clouds. But a view with some real 3d looks a lot better when viewed from an angle and it looks a lot better when you can get close. Splitting the mesh allows me to do that. There are those who would sacrifice visual effects for performance and those that would sacrifice performance for visual effects. I happen to be the latter.

cartrite

[cartrite]

After the tranparency was fixed I took some before and after shots of Don Edwards cloudmap.
This shot shows what all my cloud models looked like in OGL2 before the fix.


These shots taken while running cvs 08/30/2006 with clouds and cloud shadow on.
Basic Rendering



OGL2



I also created a couple splash screens. In case anyone wants to try them I'll post them here.





cartrite

[Don. Edwards]

It is looking great. I can't wait till you have all the files in order so you can package them up for download.

Don. Edwards

[cartrite]

Hey Don,

This fix helps a lot because I was spending most of my time trying to work around the brightnes issue I showed above.

I'm currently working on the 3D version of your clouds. With a real 3D model. The sphere model seems like it may be obsolete before it's finnished. Chris said he might enable support for normalmaps for cloudmaps. So with that in mind do you think a 3D model is needed?

I did upload the model of Katrina to the motherlode today so when thats availabe I'll start a thread in the Addons forumn. That is just the single storm in an invisible box. It fits into your cloudmap along with the one I created. The Gulf was almost empty in both caes. It should also fit into Harald Schmidt's cloudmap since I think all are using the same base map. I don't think he added any clouds in the Gulf. What troubles me is that I uploaded the wrong ssc file. I had Oblateness enabled in that file and for some that don't know better they will wonder why parts of the model dissapears. Oh well.

I sent you an PM explaining how I was going to deal with the seams issue. Did you want me to do that or should I send you the model so you can?

cartrite

[bh]

Looking bloody good...I especially like the first splash...it looks so real!

[chris]

OK. I just checked in a change that enables normal maps for cloud layers. The syntax is very straighforward. You just need to add this line to your atmosphere definition:

CloudNormalMap "your-cloud-normmap.png"

--Chris

[cartrite]

Way to go Chris and thank you.

I used a normalmap created for a model I was creating for Don Edwards cloudmap, North_summer_clouds, I came up with these screenshots. No model here. This is the normalmap enhancement checked into cvs. I suppose that now the challenge will be to come up with a good way of making normalmaps for clouds. Anyone have any ideas? I did these with the normalmap plugin for the Gimp.

I did have a problem compiling for kde. An error was produced when compiling because of the files astro.cpp and / or astro.h. I used the prior version of those files to compile this code. I got this error.

Code: Select all

kcelbookmarkowner.h:28: warning: unused parameter ?€?type?€™
celestialbrowser.cpp: In member function ?€?virtual void CelestialBrowser::slotRefresh()?€™:
celestialbrowser.cpp:130: error: no matching function for call to ?€?absToAppMag(float, double&)?€™
celestialbrowser.cpp: At global scope:
celestialbrowser.cpp:205: warning: unused parameter ?€?col?€™
celestialbrowser.cpp: In member function ?€?virtual int CelListViewItem::compare(QListViewItem*, int, bool) const?€™:
celestialbrowser.cpp:297: warning: control reaches end of non-void function
make[5]: *** [celestialbrowser.o] Error 1
make[5]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src/celestia/kde'
make[4]: *** [all-recursive] Error 1
make[4]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src/celestia/kde'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src/celestia'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia'
make: *** [all] Error 2
 

Some screenhots One shot is in basic rendering and the other in OGL2.
Cloudmap and cloudshadows are on.













cartrite

[cartrite]

refering to:

I did have a problem compiling for kde. An error was produced when compiling because of the files astro.cpp and / or astro.h. I used the prior version of those files to compile this code. I got this error.
Code:
kcelbookmarkowner.h:28: warning: unused parameter ?€?type?€™
celestialbrowser.cpp: In member function ?€?virtual void CelestialBrowser::slotRefresh()?€™:
celestialbrowser.cpp:130: error: no matching function for call to ?€?absToAppMag(float, double&)?€™
celestialbrowser.cpp: At global scope:
celestialbrowser.cpp:205: warning: unused parameter ?€?col?€™
celestialbrowser.cpp: In member function ?€?virtual int CelListViewItem::compare(QListViewItem*, int, bool) const?€™:
celestialbrowser.cpp:297: warning: control reaches end of non-void function
make[5]: *** [celestialbrowser.o] Error 1
make[5]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src/celestia/kde'
make[4]: *** [all-recursive] Error 1
make[4]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src/celestia/kde'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src/celestia'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/cartrite/bin/celestia/backup/celestia'
make: *** [all] Error 2

The Windows cvs compiled without any problems. I seen that a line refering to absToAppMag was removed from astro.cpp. I guess the celestialbrowser.cpp needs to updated as well?

cartrite

[cartrite]

What is really great is the detail that comes out in the higher resolutions.

4k with cloud shadows



8k without cloud shadows due to my card limitations.



I'm going to try a VT of the 8k map and if it works I'm going to do a 32k.

That would be really cool

cartrite

[t00fri]

cartrite,

I can't help it, but as far as I can make out from your not very discriminating images, these normal mapped "white things" do not really look like clouds to me. There is still a substantial "whipped cream" component etc.

But the general direction looks very promising, of course...Could you perhaps present some close-ups? Because that's where good clouds win over bad ones


Bye Fridger

[cartrite]

Fridger,

Will do. I'll look for some good and bad areas and zoom in.

I kind of agree with you though. There are some areas that look better than others. The problem lies with finding a method to produce a good normalmap. Areas that are mostly white with no other details can't really be normalmapped. And I'm not making these maps from signed integer height data but from grayscale cloud images. I'm still experimenting. Any Ideas on how this should be done ?

cartrite

[t00fri]

Fridger,

Will do. I'll look for some good and bad areas and zoom in.

I kind of agree with you though. There are some areas that look better than others. The problem lies with finding a method to produce a good normalmap. Areas that are mostly white with no other details can't really be normalmapped. And I'm not making these maps from signed integer height data but from grayscale cloud images. I'm still experimenting. Any Ideas on how this should be done ?

cartrite

Well I would say, like I did it years ago, when making my 2k clouds that many people used as standard clouds. At the time I bumpmapped my clouds, so the 3d structure was static not dynamic. Making the bumpmap, it is paramount to use Gaussian smearing in order to generate that typical "fluffy" impression that is so typical for clouds (and untypical for "whipped cream" )

Bye Fridger

[DonAVP]

cartrite:

I have a program called Deep Exploration. It has the ablity to make normal maps of any image I load. It allows me to input a number between 1-255 to control the bumpiness of the image (8 is the default). I used 4.

I made a quick image but was unable to upload the image on imageshack. What I did was copy the alph channel to the rgb channels in another image. Do a normal map, copy it back the 32 bit could image and us the alpha to select and delete the areas that would be transparent (hope you understand what I mean).

I posted you privately if you want me to send you this image.

Don

[selden]

Don,

If I'm understanding you correctly, your procedure starts with an 8bit bumpmap (unless you're using an image format that allows for more than 8bits in its alpha channel.) However, Celestia already converts 8bit bumpmaps to normalmaps internally. The challange is to create 24bit normalmaps that are much better than can be done with 8bit bumpmaps.

[cartrite]

I'm having the same problem with Image Shack. I guess it's the time of day. Heavy traffic?

I did another normalmap by bumping it first, on the original 8k cloud map from NASA, and it came out better. It has more contrast and less areas that are all 0xffffff. Such areas come out looking like plaster. The map I showed above was made for a 3D model.

cartrite

[selden]

Don and Fridger seem to have corrected their misunderstandings, so I've unlocked this thread.

Now get back to work on those beautiful clouds!

[cartrite]

DonAVP Wrote:

I made a quick image but was unable to upload the image on imageshack. What I did was copy the alph channel to the rgb channels in another image. Do a normal map, copy it back the 32 bit could image and us the alpha to select and delete the areas that would be transparent (hope you understand what I mean).

After seeing your normalmap had transparent areas and concerning a larger texture, isn't there a performance issue with transparencies. The normalmap for clouds also works when it has no alpha channel for opacity. the diffuse texture has that covered.

I also wanted to see what you thought of the 2k maps I sent you.

cartrite

[cartrite]

I took a couple of shots from a 32k cloud normalmap. Tile b1.









Among other things I applied a 2 pixel radius guassian blur to the bumpmap before converting to a normalmap.

This storm over the eastern US is a challenging area to normalmap. I'm not completely happy with it but it is better than the other maps I've done.



cartrite

[cartrite]

Here are a couple of 32k cloudnormalmap screenhots from a test on tile a1. I compared OGL2 with vertex program/nvidia combiners. This test was bumpmapped a little differently before creating the normalmap. I'm doing each tile a little differently to see which procedure, if any is worth doing the whole map with.










I think my FPS can be improved 500% if I deleted the tiles in my bmng normalmap folders that are not needed. The ones out in the ocean for levels 4 and 5.

cartrite

[bh]

Great job with these clouds cartrite...to me the second of your latest looks just about spot on!