3DMars update

[cartrite]

Low res hot spot models can cover the same area as high res models but with less vertices. That certainly can be done. Not sure whether it would be best to include all in one package or as separate downloads for low med or high.

EDIT: The 3dmars model I did covers a very large area. There comes a point where it is simply not worth using a model with not enough detail when you under sample the data. In my opinion, the 3dmars med res areas where as scaled down as I could get. This new approach will have no 3d details for most of the planet. Just the hot pots. The rest of the planet would use a normalmap.

Danielj,
Just to remind you, I first created that 3dmars model with a ge fx5500 card, a k7 900mb processor and 512 mbs of ram. The enhancements for version 1.5 or the texture size you are using is what is probably bogging down your frame rates. Try to scale them down with an image editor.

End Edit
cartrite

[ElChristou]

UNBELIVEABLY that the 8600 GTS gives at least more 10 fps than the 7600 GT.In my system,it runs at the ridiculous frame rate of 12-15 fps and the 2 GB RAM didn??t help in anything.It appears that I increase my memory to NOTHING.It??s an ABSURD that I should change my video card.Do you realize that very few users have a computer powerful enough to run this cumbersome model?Only Andrea and a few more...

Again...

AND WHAT? BETTER DOING NOTHING?
That way not only you but no one would be able to enjoy those models? Is that what you mean?

(BTW, daniel, I got 256Mo of ram... so if your 2Go don't serve, you can always send them to me! )

[ElChristou]

..That certainly can be done. Not sure whether it would be best to include all in one package or as separate downloads for low med or high.

Well if it's possible, I suppose the several hot spot res should be release separately... (Perhaps the Low res global mesh could be released with low res hot spot, then the med and high separately (with descriptions and shots on your site to let the user decide what best for him...)...

[cartrite]

..That certainly can be done. Not sure whether it would be best to include all in one package or as separate downloads for low med or high.

Well if it's possible, I suppose the several hot spot res should be release separately... (Perhaps the Low res global mesh could be released with low res hot spot, then the med and high separately (with descriptions and shots on your site to let the user decide what best for him...)...

I am just wondering if it would be possible to have a base mesh. Then all the high res regions could be downloaded as one sees fit and have them "fit" into the model. Something the way virtual textures work.
cartrite

[ElChristou]

I think IF it's feasible, it would be more convenient to make each spot independent and in several res. This way it should give more flexibility to the end user depending of his config (perhaps one would be able to load only one hot spot in highres but not one mesh with all the highres spots...)...

[cartrite]

I think the only way to make each hot spot independent is though code. Dynamic loading. The base mesh would have all vertices erased between minx and maxx and miny and maxy of the model. Then the model would take the place of the erased vertices. Perhaps a model loading module called by (or built into) Celestia through Lua scripts or something?

As for now, I create a transition zone with more vertices so I can get the base mesh "close" to the z values of the high res mesh. Then the hires model sits in the middle of the zone. Since every model would need a different zone, the only way I can think of is make a separate model for only that hot spot. And then I could make a separate model for each resolution. These images try to show why you need extra vertices around the model and why the z vales need adjusting.





There are "floating" models that could be loaded now slightly above the planet. This would work fine except if the mesh were below the planets surface radius. As is the case with the Vallesmarineris. Thats why I abandoned that approach.

Anyhow the easiest thing to do for now is to make a separate mesh for each area.

cartrite

[danielj]

HOW CAN I DO IT?The 3d Mars I have downloaded have 9.64 MB,but have only a model.The textures are not SEPARATED...is there another download that I didn??t know?

Low res hot spot models can cover the same area as high res models but with less vertices. That certainly can be done. Not sure whether it would be best to include all in one package or as separate downloads for low med or high.

EDIT: The 3dmars model I did covers a very large area. There comes a point where it is simply not worth using a model with not enough detail when you under sample the data. In my opinion, the 3dmars med res areas where as scaled down as I could get. This new approach will have no 3d details for most of the planet. Just the hot pots. The rest of the planet would use a normalmap.

Danielj,
Just to remind you, I first created that 3dmars model with a ge fx5500 card, a k7 900mb processor and 512 mbs of ram. The enhancements for version 1.5 or the texture size you are using is what is probably bogging down your frame rates. Try to scale them down with an image editor.

End Edit
cartrite

[danielj]

HOW CAN I DO IT?The 3d Mars I have downloaded have 9.64 MB,but have only a model.The textures are not SEPARATED...is there another download that I didn??t know?
Anyway,I download the lores pack,but I can only get 12-15 fps!

Low res hot spot models can cover the same area as high res models but with less vertices. That certainly can be done. Not sure whether it would be best to include all in one package or as separate downloads for low med or high.

EDIT: The 3dmars model I did covers a very large area. There comes a point where it is simply not worth using a model with not enough detail when you under sample the data. In my opinion, the 3dmars med res areas where as scaled down as I could get. This new approach will have no 3d details for most of the planet. Just the hot pots. The rest of the planet would use a normalmap.

Danielj,
Just to remind you, I first created that 3dmars model with a ge fx5500 card, a k7 900mb processor and 512 mbs of ram. The enhancements for version 1.5 or the texture size you are using is what is probably bogging down your frame rates. Try to scale them down with an image editor.

End Edit
cartrite

[cartrite]

danielj,
Assuming you got the 3dmars model from the Celestia Motherlode, I can't see that model getting anything less than 45 fps with a 7000 series card. True if you can't put more than a 4k texture on it then a planet wide 4k in not that good resolution. Version 2 of the model can be found here. I thought you had this model. If this model runs too slowly, it shouldn't, I was saying you could scale down the 4k textures that comes with it to 2k textures. Pixel for pixel thats still a lot bigger than a planet wide 4k for the areas covered.

PLEASE don't confuse the model at my site with the model in these last few pages. The model shown above that created the images on this page and page 4 is experimental and has not been released to the public yet.

cartrite

[dirkpitt]

I think the only way to make each hot spot independent is though code. Dynamic loading. The base mesh would have all vertices erased between minx and maxx and miny and maxy of the model. Then the model would take the place of the erased vertices. Perhaps a model loading module called by (or built into) Celestia through Lua scripts or something?

Aren't models in Celestia already dynamically loaded only when visible? For example, the base (low res) mesh is always loaded, but if the high-res Valles Marineris model is hidden on the other side of the planet it won't be loaded, etc.

[cartrite]

I think your right dirkpitt, but that's not what I meant. Sorry.

I think to get this to work, the base model would need to be modified in real time. For example: The coordinates for the first model I made were as follows:
Min x = -0.409057
Max x = -0.395141
Min y = -0.058105
Max y = -0.030762
All the coordinates that fell between those ranges would be erased from the base mesh and the model would supply the new coordinates. The base mesh would also reshape itself to fit up to the model by recalculating the values for z. After the user changed his or her mind the model would be erased from memory and either the base mesh would replace the old values or another model would be chosen.

That is kind of what I'm doing now. Except I'm doing it permanently by building the model.

If that ever happens it would be sometime in the distant future. For now the only way to create a lot of models, is to create one for each area of interest. The files are just tooooo big. That is, when the resolution is set so high. Keep in mind that the textures that I have been displaying are only about 6 meters per pixel. 25 mpp's for the Mex model. I have data that is in the .3 meters (1 foot) per pixel range. To render those textures in 3d would take a highly detailed dem.

cartrite

[danielj]

The old version didn??t work at all.I unzip it to Extras in Celestia 1.5.0 pre3,but Mars was messed up.The planet become featurless,only with the Valle Marineris and the Tharsis Mons,and for most part,grey.And when I tried to change the texture to a 4k one from Don Edwards,Celestia simply crashed.I will try with Celestia 1.4.1!

danielj,
Assuming you got the 3dmars model from the Celestia Motherlode, I can't see that model getting anything less than 45 fps with a 7000 series card. True if you can't put more than a 4k texture on it then a planet wide 4k in not that good resolution. Version 2 of the model can be found here. I thought you had this model. If this model runs too slowly, it shouldn't, I was saying you could scale down the 4k textures that comes with it to 2k textures. Pixel for pixel thats still a lot bigger than a planet wide 4k for the areas covered.

PLEASE don't confuse the model at my site with the model in these last few pages. The model shown above that created the images on this page and page 4 is experimental and has not been released to the public yet.

cartrite

[cartrite]

The old version didn??t work at all.I unzip it to Extras in Celestia 1.5.0 pre3,but Mars was messed up.The planet become featurless,only with the Valle Marineris and the Tharsis Mons,and for most part,grey.And when I tried to change the texture to a 4k one from Don Edwards,Celestia simply crashed.I will try with Celestia 1.4.1!

danielj,
Assuming you got the 3dmars model from the Celestia Motherlode, I can't see that model getting anything less than 45 fps with a 7000 series card. True if you can't put more than a 4k texture on it then a planet wide 4k in not that good resolution. Version 2 of the model can be found here. I thought you had this model. If this model runs too slowly, it shouldn't, I was saying you could scale down the 4k textures that comes with it to 2k textures. Pixel for pixel thats still a lot bigger than a planet wide 4k for the areas covered.

PLEASE don't confuse the model at my site with the model in these last few pages. The model shown above that created the images on this page and page 4 is experimental and has not been released to the public yet.

cartrite

Here is a file list from 3dmars/textures/medres/

Code: Select all

mons0.jpg
mons1.jpg
mons2.jpg
mons3.jpg
mons-norm0.jpg
mons-norm1.jpg
mons-norm2.jpg
mons-norm3.jpg
spec.jpg
tex-spec.jpg
valles0.jpg
valles1.jpg
valles2.jpg
valles3.jpg


Celestia comes with a texture called mars.jpg
If you did not rename that file the model will use it.

Code: Select all

#celmodel__ascii



material

diffuse 0.878431 0.878431 0.878431

specular 0.498039 0.498039 0.54902

texture0 "mars.*"

normalmap "norm.*"

specularmap "spec.*"

end_material



Notice that the model looks for a file called norm.something.
That file used to be in the folder but I think I removed it to decrease the file size. I forgot to mention that in the instructions. If you have a normalmap of mars, copy it and rename the copy norm.something and replace something with whatever file format the normalmap is in. Example: If you have a normalmap of mars called mars-normalmap.png, name the copy norm.png. mars.* and norm.* do not have to be in the model directory. They will be found if they are in the celestia/textures/medres folder too. After you get that right, the features like the normalmap and specular highlights only work in ogl2. If you had a file in your Celestia folder called mars.png or mars.jpg or something, you should have seen it. Apparently you don't. As for Celestia 1.4.1. The model was not made for it so they will not work. It's enhanced features were meant for Celestia 1.5 and will only work with Celestia 1.5, or the cvs version.

cartrite

[ElChristou]

I think the only way to make each hot spot independent is though code. Dynamic loading.

That would be nice, but seems it's not for tomorrow... In the meantime, a classical addon structure is not enough? Each spot would be shipped with is own ssc placing it to the right place in the puzzle. Then one would have to choose what files to put in the extra folder... (of course, one would have to remove by hand the ssc of the low res before using a high res one...)

[cartrite]

I think the only way to make each hot spot independent is though code. Dynamic loading.

Each spot would be shipped with is own ssc placing it to the right place in the puzzle. Then one would have to choose what files to put in the extra folder... (of course, one would have to remove by hand the ssc of the low res before using a high res one...)

I gave up trying to place a model in the right place with a ssc file long ago. It was really frustrating and confusing. Fortunately for me I learned that by orienting the model correctly before exporting it to cmod, Celestia places it in the right place without ssc file adjustments.

cartrite

[ElChristou]

I gave up trying to place a model in the right place with a ssc file long ago. It was really frustrating and confusing. Fortunately for me I learned that by orienting the model correctly before exporting it to cmod, Celestia places it in the right place without ssc file adjustments.

cartrite

Agree, it's what I was thinking in, just a ssc to call the model and define it's root...

[cartrite]

Here is a series of screen shots that attempts to show the process I do to orient the model.

The first shot shows a plane that was subdivided. The 2nd shows the lowres mesh that was already created.



The next 2 show the lowres mesh "hole" and the lowres mesh in the "hole"
The area highlighted in yellow was used to come up with the minimum and maximum x y z coordinates to export the MEX dem to a waveobject .obj file. The MEX dem also fit in this whole after importing it into Blender.



The next 2 show before and after the z coordinates were modified. These z coordinates were modified with a MOLA height map.



This next shot shows the whole plane before running the z-fold script.



Next, this shot shows the model after running z-fold.



Then the whole planet.



The last thing I do to orient the meshes is to rotate the x axis -90 degrees and the y axis 180 degrees. Then Celestia draws the model in the correct positions. Note: At this point I can delete the planet and just export the lowres mesh and it will still be placed in the correct location. If the mesh were above the planet's radius, everything would work fine. But since this mesh is in a canyon it is underneath the planet and would not be seen if a "hole" were not created.

cartrite

[ElChristou]

After running the z-fold, we can see some "leaks" around the high res model; are you able to correct this?

[cartrite]

After running the z-fold, we can see some "leaks" around the high res model; are you able to correct this?

That where the WORK starts. It's very tedious. I have been practicing moving the vertices from the planetary plane in the z direction to close the gaps but I think it may be impossible to do manually. There are just too may of them. I did see that the more verticies there are in the transition zone the less leaks there are.

I was thinking of exporting the coords from the perimeter of the mesh that fits into the whole and the perimeter around the whole to see if I could use that data to modify the z coords around the whole after running z-fold. I'm still trying to think about how this could be done with a python script.

Here are a couple of shots of the actual MEX dem that was scaled down 50%.
I'm thinking I may get away with scaling it down to 25% of the original. Anything less than that, well I might as well use mola data. That would be a point per about 800 meters or so for this mesh. Not that there is anything wrong with mola data, but mola data has a maximum resolution of 128 pixels per degree and I trying to model a texture that has about 9000 pixels per degree.



cartrite

[ElChristou]

That where the WORK starts. It's very tedious. I have been practicing moving the vertices from the planetary plane in the z direction to close the gaps but I think it may be impossible to do manually. There are just too may of them. I did see that the more verticies there are in the transition zone the less leaks there are.

I was thinking of exporting the coords from the perimeter of the mesh that fits into the whole and the perimeter around the whole to see if I could use that data to modify the z coords around the whole after running z-fold. I'm still trying to think about how this could be done with a python script.

I suppose those leaks appear because there is 3 meshes (global low, transition zone and Highres zone)... what if you weld them before the z-fold? Then there must be a way to separate them again after being folded...