New Celestia-1.4.0pre-FT1.1 Version for Download

[jdou]

No bug see my post above

[t00fri]

Jdou,

thanks, I always prefer NO bugs over bugs...
Cheers,
Bye Fridger

[ElChristou]

Are the actual templates only the "heart" of the galaxies????

[t00fri]

Chris,

certainly NOT. The templates are precisely scaled to the dimensions the galaxies have according to the professional catalog data. The light curves are measured across true-color galaxy images and reconstructed within Celestia.

What you did above is to compare "apples" with "peers" !
You compared a super hires image from Hubble on a completely different image scale with a lores Celestia template that was even not correctly illuminated (or your monitor badly adjusted for brightness and contrast).

Here is how M100 should look like. Since in deepsky.dsc it's SBc, our template approximates it's shape as displayed below. There is the following problem though: Simbad says M 100 has Hubble type Sc, while the revised NGC/IC catalog atributes it to SBbc (same type as the Milky Way) which our simple Hubble type conversion maps into SBc (like our Milky Way). So there is clearly room for improvement here by introducing subclasses to the Hubble type classification.

Repeated again: our approach can NEVER get individual galaxies perfectly right, but rather the task is to get 10000+ galaxies mostly right. It is completely besides the point to confront the results for a particular galaxy with a giant resolution Hubble photo!




Bye Fridger

[Cham]

Fridger,

the SBc templates are badly done. They are VERY ugly and doesn't feel natural at all. IMHO, the previous versions were better.

[t00fri]

Fridger,

the SBc templates are badly done. They are VERY ugly and doesn't feel natural at all. IMHO, the previous versions were better.

Cham,

sorry, but I disagree. Here is the image from NGC 1300 that is a classical SBc spiral and served as input image for the vectorizer, i.e the above SBc spiral. You will recognize the similarity.



It is besides the philosophy to try and get back to the old hand-crafting idea referring to individual galaxies. Of course then you can do better. But you cannot do 10000 galaxies better! You cannot do 10000 galaxies at all, otherwise!. I hope in your critics you took also into account that we only can afford about 200KB of size per template! Why don't you try to get something better done for NGC 1300 that is NOT bigger than 200KB!?

You can also see from this high-quality image that the light curve is NOT as you previously claimed it had to be. In the center it's NOT a glaring patch of light as you requested.
Our template light curve is precisely measured from the above image in natural 2500x2500 pixel size!

The point of the present approach is rather to concentrate on characteristic Hubble shapes like the above SBc such that with a glimpse of the eye one can recognize the type. We can only hope to make good "caricatures" of the galaxies in question! It can only be the systematic consistency with reality that we may achieve and which will make the present scheme useful and unique.


Bye Fridger

[Cham]

Fridger,

I disagree with you here. Here's why :

The template idea is to reproduce the standard mean shape of a given galactic type. Why then use a very particular galaxy (NGC 1300) to represent them all ? It SHOULD NOT be used as a typical representation. You need to design a MEAN, TYPICAL representation of that galactic type, not of NGC 1300. A kind of AVERAGE over ALL barred galaxies, not a special specimen.

I find that the NGC 1300, as seen on this picture, is VERY different that the template itself. Not much ressemblance here, really, with the extremelly ugly template used. The NGC 1300 is a beautifull specimen, no doubt about it. However, the template is a very poor representation of it, and I feel it's a shame that this model could become the final template in the official version.

If you look carefully at the template, you'll notice a large disymetry in the arms which ISN'T apparent at all on the real pictures. Like I said, this template needs some corrections.

[Cham]

Here's what I mean :

[t00fri]

Fridger,

I disagree with you here. Here's why :

The template idea is to reproduce the standard mean shape of a given galactic type. Why then use a very particular galaxy (NGC 1300) to represent them all ? It SHOULD NOT be used as a typical representation. You need to design a MEAN, TYPICAL representation of that galactic type. A kind of AVERAGE over ALL barred galaxies, not a special specimen.

I find that the NGC 1300, as seen on this picture, is VERY different that the template itself. Not much ressemblance here, really, with the extremelly ugly template used. The NGC 1300 is a beautifull specimen, no doubt about it. However, the template is a very poor representation of it, and I feel it's a shame that this model could become the final template in the official version.

If you look carefully at the template, you'll notice a large disymetry in the arms which ISN'T apparent at all on the real pictures. Like I said, this template needs corrections.

Cham,

first of all, you forgot that the image of NGC 1300 has to be unfolded by computer, since the view is NOT face on. That's where the "distortions" come from that you inadequately critisize. If I had not done that, the final appearance of type SBc galaxies would not be to size.

I told you: the present template is NOT handi-crafted. It's computer generated with Toti's bmp2pts code. I did not even touch it. All you can adjust are 2 parameters for the vector conversion. Any more fancy conversion would blow the template size up enormously!

Before going on, please try and design a shape within 128x128 pixels (!) that is not more than 200KB yet has all the details that you would like it to have.

You said incorrectly that the above NGC 1300 shape was not a typical SBc shape. That's clearly incorrect. Let me just state that I spent weeks experimenting with the templates and I am certainly NOT a beginner in image manipulation. It's just NOT easy to do better given the constraints we have. You can impossibly average over many SBc galaxies to generate the template in practice.
You must be lucky to have one good one available...

Bye Fridger

[Cham]

Fridger,

the template we had some weeks (months?) ago was better, IMO. It was a better representation of most barred spirals, and I'm pretty sure it was satisfying your constraints.

The problem with this new template is its apparent asymetry. It just needs some corrections on the arms.

[t00fri]

Here's what I mean :

Toti's program makes points where there should be points according to a sensible underlying algorithm. It leaves them away if certain intensity and random criteria are not met.

As soon as I lower the sensitivity threshold sufficiently, then all points of the ngc 1300 image will be mapped, with the result that the image is better, i.e closer to the original, but the template size is 10-20 times as large!

We may hope that Toti's Mark II version will be some improvement, but personally, I doubt it.

Bye Fridger

[Cham]

Toti's program makes points where there should be points according to a sensible underlying algorithm. It leaves them away if certain intensity and random criteria are not met.

We may hope that Toti's Mark II version will be some improvement, but personally, I doubt it.

Bye Fridger

Then Toti's program isn't adequate for this galaxy type. It needs to make some kind of average. Using only one galaxy sample isn't enough to make a sufficient "average" or "typical" representation. Apparently, its program is amplifying some features of the specimen used, which leds to this asymetry I was talking about. The result isn't satisfying. Geez, it even looks like an "amateur" work ! (I'm talking about the template itself, NOT Toti's work !).

[BlindedByTheLight]

Toti's program makes points where there should be points according to a sensible underlying algorithm. It leaves them away if certain intensity and random criteria are not met.

As soon as I lower the sensitivity threshold sufficiently, then all points of the ngc 1300 image will be mapped, with the result that the image is better, i.e closer to the original, but the template size is 10-20 times as large!

We may hope that Toti's Mark II version will be some improvement, but personally, I doubt it.

Bye Fridger

Got a question... partly for my own edification, partly because I like to weigh into conversations that are clearly far and away over my level of expertise...

1) The Process: do I have it correct to say that Celesia uses a small number of what we are calling "templates" (say 1, 2, 3 or so) that are representative of ALL the galaxies to be displayed?

2) And those templates are simply some kind of graphics file on the computer? (where are they if that's correct - I couldnt' find them)

3) Then Celestia just "repeats" those five or ten graphics files in the right location (I'm guessing maybe with tweaks for brightness and/or color) over and over and over per the specs in the 10,000+ catalog?

4) Those templates (graphics files) are created with an algorithmic program made by Toti... which samples points from real galaxies to come up with the "template"?

If all that is true (and, at this point, that's a stretch)... is it not possible then, after the "templates" are "made", to go back and artistically tweak them so that they better approximate reality?

Sometimes a direct sampling of what is REALLY there... doesn't look like what is really there... without some additional tweaking, 'na mean?

I'm not suggesting that all 10,000+ be tweaked... but, if I understand the process right... aren't all 10,000 based on just a few templates?

Yatta!

[Cham]

Sometimes a direct sampling of what is REALLY there... doesn't look like what is really there... without some additional tweaking, 'na mean?

Yatta!

That's exactly what I'm trying to say. The sampling for the barred spirals isn't right. It most be corrected, tweaked in some way. Toti's program isn't "smart" enough to do this. Very often, we must "cheat" with reality representations to get a better picture of the reality.

[t00fri]

Fridger,

the template we had some weeks (months?) ago was better, IMO. It was a better representation of most barred spirals, and I'm pretty sure it was satisfying your constraints.

The problem with this new template is its apparent asymetry. It just needs some corrections on the arms.

Again, you forget that the present converter does NOT allow to manipulate individual points! All you can do is a most tedious procedure by pre-treating the input .bmp model in contrast and brightness in certain areas. Then after each modification and re-conversion to .pts, you got to copy the new model into Celestia, start it and inspect it...Moreover there is randomness involved. So the thickening on the rhs arm of the template was such a random effect. I just can hardly do anything about it at this time.

The present SBc template you dislike cost me fare above 100 such cycles. You can easily estimate the large required time/template...

Bye Fridger

[Cham]

Fridger,

then Toti's program isn't good enough. It isn't adequate (or not "smart" enough). You need something more sophisticated to model a "typical" galaxy.

[BlindedByTheLight]

Again, you forget that the present converter does NOT allow to manipulate individual points! All you can do is a most tedious procedure by pre-treating the input .bmp model in contrast and brightness in certain areas. Then after each modification and re-conversion to .pts, you got to copy the new model into Celestia, start it and inspect it...Moreover there is randomness involved. So the thickening on the rhs arm of the template was such a random effect. I just can hardly do anything about it at this time.

The present SBc template you dislike cost me fare above 100 such cycles. You can easily estimate the large required time/template...

Bye Fridger

So... if I may interject as moderator... it sounds like Fridger agrees it would be lovely and nice to tweak the models, but there just aren't the software tools available to do so in a time-efficient manner? In other words, these are NOT 2D pictures that can just be diddled with in Photoshop... they are 3D models with thousands of little points that are not easily to adjust and QC (with current tools)? Do I have that right?

Still curious about the process though (as written in my prior post).

[BlindedByTheLight]

I would also like to highlight Fridger's comment about time...

It seems to me there are two issues... one, setting up a paradigm in Celestia to even make such a thing as using the catalog to display galaxies POSSIBLE in the first place.

And two, making sure the templates themselves are graphically representative of the galaxies.

Those are two very different tasks. And it seems to me that each takes quite a bit of resources. So, thus far, Fridger and Toti have chosen to make sure ONE is working right and also that TWO is working as well though TWO, all agree, could be better.

But to make it better, at this time, would take away resources from ONE - something F&T have chosen against.

That about sum it up?

[Cham]

Here are the templates used for M 100 from Celestia 1.3.2 up to the actual FT 1.1 version. These are FRONT views and I did NOT adjusted the brightness. Please, don't compare the luminosity, it's isn't what I'm talking about here :

Celestia 1.3.2 :


Celestia 1.4 (Dirkpitt version on OS X) :


Celestia FT 1.1 :


Personally, I find the third view is absolutly HORRIBLE ! It looks like tilted in some way, grossly deformed. It's CRUDE ! The first one was much better. I think the best model should be something between the first and the second models, NOT as a representation of M 100 alone, but as a representation of all barred spirals of that type.

[Malenfant]

I notice that all of these templates seem to have the core being the same brightness as the arms. Yet in all the photos I've seen of galacies, the core is much brighter and certainly doesn't look like its composed of lots of little dots. Yes, I know technically a galaxy is composed of lots of little dots (ie stars), but the templates don't give the right impression IMO - the dots are too large and sparse and smudgy.

FWIW I do agree with Cham, the shape in the FT version does look less pleasing than in the previous ones - it looks horribly asymmetric and bent out of shape and well... amateurish in appearance. It's ugly, IMO.

(note to Fridger. Please don't take these criticisms personally. This is good work you and Toti have done - we're just trying to point out where the templates could be improved).