Museum wall

Started by Dune, August 26, 2013, 03:21:28 AM

Previous topic - Next topic

Dune

I'm about to embark on quite a demanding task; making a 150 x 10 foot wall for a museum, all TG3. It's supposed to be one terrain, extending from right to left, spanning 250.000 years.
So..... thinking about this gives me the creeps already, and I would like to have any input by you guys. Anyone have any experience with this?
Luckily there are gaps, and it doesn't need to be one huge file. But still, they have to merge into each other, which I will do in PS. But perhaps PS can't handle a huge file. And how about cutting it in pieces and then stitching the non-merged bits together? The cut bits will fit again exactly anyway. But what will a printer need, one huge file, or can he work with the fragments? It will probably be printed on something flexible to accommodate wall curves.

And how large should I render? 30-50px/inch, billboard quality I guess. How many inch in a foot? Why don't we all just use the metric system? I'll figure this out, no problem.

But I also saw that if I take a render ratio of 1000x300 for a chunk, and I need a less wide chunk next to that, say 600, but same height 300, the whole POV changes. What would be better, using all the same chunks? Looks like it. I can divide the whole wall in 5 chunks, or 10. The latter is easier to work in of course, but needs more stitching work. But it seems another ratio changes the perception of the terrain......

Speaking about fitting chunks together: I can't really change the sun, as it would be some sort of cylindrical view, flattened to this wall. If I change the sun from one chunk to the next to keep light always from one angle, you get shadow discrepancies. So, one sun. There's bound to be an area with back light and/or one looking into the sun.

Another difficulty: it's a hilly terrain, but near the floor of the museum they want grass and stuff. like you're real close, so not an aerial view. That can be done, but I like to see into the distance as well (though not necessarily everywhere). So the camera should be on a hill or you look against the first hill and no further. But if you have the camera on a hill, 1.6 meters above ground level of that hill (which may be 40m high), you have a valley in front of you, and you can't really see the bottom, or you'd have to point the camera down. But if you point the camera down, you won't have the horizon at 1.6 meters above actual floor level. If you use wide-angle, there's too much distortion. So I guess a normal lens of 60 (default) or maybe 50 would be best to use.

Again another problem is to make it believable. The right hand side of section 5 needs to be a forest, but I only got 2-3 meters or so. Then you get a meter of tundra, a meter of ice age, another meter of tundra, then wood again. Wow.

If anyone has some bright ideas, I would greatly appreciate them.

Oh yes, and these are just tests, ground and veggies need a lot of work, and by coincidence 2 instances of a tree are rotated exactly alike.

kaedorg

Hi Ulco

1 foot = 12 inches
1 inch = 2.54 cm

I have printed one of my tg3 view and use a render of 300px/inch or 120px/cm
To give you this example i printed a 60cmx40cm so i used a render 7200x4800 dpi

Hope it could help you

David

mhaze

It will be important to speak to the printer to see what he can handle.

Mahnmut

Hi,
I know that this is only an early test, and I think it´s a good start.
Just some ideas about some of your questions:
-If you need panels with different proportions, but the same POV , can you simply crop them?
-One meter per "age" isnt much, ´but if you concentrate on the midground (except perhaps for the wooded times) you have a lot more space for the transition, an iceage would in the foreground be no more than a meter wide lump of snow, but in the midground there is room for mammoths.
- I am not sure if shadow discrepancies would be so bad. If you can integrate your looking into the sun artistically, then it´s fine. But in my humble opinion, shadows moving slightly (you don´t have to maintain the sunlight at exactly one angle relatively) in a panorama that represents millenia could also have an interesting effect. just look what works better, I just want to say shadow discrepancies aren´a knockout, I think.
that´s myideas, Ihope they help a little.
best Regards,
Jan

choronr

Just your writing up the challenges that face you shows one that you already have a good focus on what you are about to encounter.

A very good converter: http://www.worldwidemetric.com/measurements.html

Dune

#5
The hills have to be much lower, which has its advantages and disadvantages. Less interesting, less distance possible, but you can look into 'valleys' now. Another test, just checking if I can get a medieval road and a Roman road in without conflict. Found that if I turned the camera for the other sections I get the same road in the ice age  :-\ Good for the reindeer.
I also found that I have to make sections of 3:1, or I will get to the same point again while turning the camera. If I make it 2:1, I have to turn the cam 50 degrees, same as with a 3:1 ratio, but have to make more turns to fill the 50 meters (80 m. in fact!), thus ending up more than 360 degrees around. And I have to turn, or the alignment of objects won't stand. You can't just move the cam aside and turn. Pretty hard nut to crack..... 

detail 0.5, AA 5, 1+ hour.

@Jan: It would be great if I could use just one POV, and crop render different parts (or render as a whole), but you can't work properly in a very wide ratio (50:3) render frame. Try it and you'll see that you can't zoom in. So I have to use less ratio, and turn the camera. I also agree about the ice not extending to the front (I already had a distance shader mixed with the camera mask, but turned it off while doing these tests). One meter of snow isn't working, especially not with the tree shadows across the ice age snow. Shadow discrepancies may indeed be not too bad. I can always blur them together in post. Unless they're very distinct, like form buildings. But they will be far out anyway. But I think I can get away with a steady sun.

Thanks guys.

mhaze

This is far more than I would ever take on.  Early results are very impressive and I look forward to more.

TheBadger

#7
Should the sections not represent the duration of the age? That is, If from one end of the printed work to the other is 100%. Than what percent of the time periods represented was the ice age?

Also, in terms of construct, it feels a little like you are attempting to work this out in the software?
Why not attempt to represent the possibilities by "thumbnail" on paper?

For example, the composition is going to be of great importance, not just in the usual way, but also in terms of representing history artistically or scientifically (literally). So, if (again for example), the ice age represents 20% of the total amount of time being covered, and say that, that time compositionally falls center frame. Than what does it look like with a lot of white in the center, flanked (presumably) by lots of greens?

Since you already know the dimensions/aspect ratio of the final. Than simply print out a scale box of the frame and on a bunch of paper and start coloring in ideas. Not details like hills and trees, but color information by percentage... The colors being representative of both period (non literal sectioning) and also literally the colors you would see at the time on the landscape (plants, snow, so on)

It sounds like your spending a lot of time just trying to figure the canvas out (so to speak) But I again would repeat what I said in the other thread on this topic a while back. use photo techniques:
http://www.aaronpriestphoto.com/pano/2013-04-13_Spring_or_Winter/ zoom in and see what I mean
You could simply process the frames flat with no distortion to hang on a wall in the traditional way.

Same could have been done with this:
http://fc01.deviantart.net/fs71/i/2011/150/2/3/panorama_museum_1453_by_lapinaru-d3hn8em.jpg
Work in 180-360 degrees. and stitch them.
(Autopano pro) You will have to render what would be the ground at your feet, and lots of the sky to get a flat pano (no distortion) that would all be cut out in the end though.

p.S. edit
like this http://photography-on-the.net/forum/showthread.php?t=1287247 but with even less distortion ;)
It has been eaten.

bigben

I'd agree with a few of the suggestions here.
Definitely start with the printer to see how it will be printed.  That will determine how you will have to slice the image. If you can produce a single render for each panel that should be the simplest

Doing it as a 360° panorama would provide a range of lighting angles that could be used to good effect.  If you're printing a number of vertical panels you wouldn't need to stitch the images together. Just render each panel and then re-project it to a cylindrical projection. For your given proportions, that will result in a vertical fov of 23.7°. You'd need to render a little higher than this to allow for re-projection, how much depends on the horizontal fov of each image.  If you want help determining the required vertical fov for this just drop me an email when you know how you're slicing the image.

j meyer

What are the gaps you mention in the first post?

bigben

Quote from: bigben on August 27, 2013, 09:23:23 AM
.... Doing it as a 360° panorama would provide a range of lighting angles that could be used to good effect.  If you're printing a number of vertical panels you wouldn't need to stitch the images together. Just render each panel and then re-project it to a cylindrical projection. For your given proportions, that will result in a vertical fov of 23.7°.

Silly me  ;)  Of course with a cylindrical projection you could do whatever horizontal fov you wanted to since each piece will just butt up against the previous one.  You could for example, go to 540° and still have one sun
You would need to toggle the visibility of overlapping parts as you rendered or it might be easier to split it into multiple TGDs covering 180° so that you can duplicate part of the scene in each that consitutes the transition and then change everything else.  You could even do multiple revolutions, obscuring additional occurences of the sun by clouds.  The wider you go horizontally, the more vertical fov you can use to provide a better perspective with foreground detail (difficult to do with only 24° vertically)

If you don't have much space for forests, possibly just use  one or two large trees at the back and gradually increasing density of smaller trees in front. The size and type of trees used should give enough of an indication.

Dune

I appreciate your thoughts on this, guys. Michael: the archaeologist already put all detail on paper, eras, amount of wood, details he wants, so I got the layout in front of me. My main problem is to make the terrain continue and populate and color it as the eras progress. A cylinder is the only thing that works, or shrub or hillock or road in front won't line up when stitching. I don't think I need to stitch them all, as there is some overlap and gaps here and there (a door, a little nook going around a corner), but the terrain more or less needs to continue.
I plan to mask eras by a camera driven gradient mask, so you get 'slices of cake' projecting outward. This could all be one tgd. But as you can see in the panorama photo you linked to, Michael, the ground area will be quite small, as you move around on that one tiny spot with your 'camera'. It's maybe 10m wide. So you can't really color the eras in the grass at your feet, so to speak. But as you walk along that wall in the museum, it would be nice to see different vegetation at your feet at each section of the 78m (!) wall. A more or less aerial view without grass at your feet would be easier, but they want grass at your feet (like your walking along a grassy lane looking at the ages). I think that's my main concern, but I think it's unsolvable unless with a lot fiddling in photoshop. It will be more like standing on a grassy hill looking around at the ages, but then on a flat wall, not like in the renderosity diorama. So a bit of a discrepancy to be overcome.
So that's why I have to figure this out in TG3, it's my only option. That's my camera shooting pictures.

By the way, the ice age is the hard part as that's only a few meters, flanked by woods, as you said, it's only 5% of the wall. The rest is more gradual, wooded, less wooded, etc. I agree using a few (smaller) trees in front and making woods denser in the distance, or you wouldn't be able to the distance.

What do you mean (Bigben) by 'the wider you go horizontally, the more vertical fov you can use, etc'? If I change the camera angle from 60 to 80, e.g. you might get angled trees at the edges, which will be harder to stitch. I also found that if I change the ratio, it won't just cut off a piece of left and right, but also from top and bottom. Which is what I don't want. It condenses the whole terrain, as you can see from these screengrabs. So I can't use different ratios for different sections or they won't fit together. Or am I not clear in the head?

TheBadger

Ok Ulco. I was going to post this before I saw your response. But then I thought I should post it anyway, because it may still at least provide some creative stimulation.... Of course it was designed to be (and was) a moving picture. But nonetheless it demonstrates a creative solution used in a very famous film.

[attach=1]

[attach=2]

[attach=3]

Of course the ideas I was playing with would be terribly render heavy. You necessarily would need to overlap. So a one render solution is ideal.
It has been eaten.

bigben

Quote from: Dune on August 28, 2013, 03:26:38 AM
What do you mean (Bigben) by 'the wider you go horizontally, the more vertical fov you can use, etc'? If I change the camera angle from 60 to 80, e.g. you might get angled trees at the edges, which will be harder to stitch. I also found that if I change the ratio, it won't just cut off a piece of left and right, but also from top and bottom. Which is what I don't want. It condenses the whole terrain, as you can see from these screengrabs. So I can't use different ratios for different sections or they won't fit together. Or am I not clear in the head?

If I understand your approach, you're looking at doing this in a single camera view(?)  That limits your options for vertical fov.  What I'm referring to is using several camera views and converting each render to a cylindrical projection. If your camera is horizontal (eliminates the need for stitching), you can just keep rotating your camera around to build up the scene. So if you're looking at roughly 8:1, for 40° horizontal fov you have only ~5° vertically.  If you use 2 x 40° views you then need ~10° vertical fov to keep the ratio and so on.  I'd work the problem in reverse.  Find the vertical fov that gives you the depth perspective that you're looking for and then calculate the horizontal fov of a cylindrical image to match. If it's less than 360° then you're laughing. If it's greater than 360° it's still possible (but challenging), and I can explain that if needed.

For the mask I'd use a vector in the xy plane from the texture xyz to the camera position and then create the masks for specific bearings. That makes the mask independent of camera movement

bigben

Just a wee clarification. It's likely that you may want to tilt the canera upwards, which is ok for determining the composition, but for this approach you'd need to render with a pitch of 0 and crop from the bottom. Most people don't push cylindrical images beyond 120 deg. vertically but that allows for a 90 deg vertical fov tilted up at 30 deg.  All you need to do is play around to get the vertical perspective you want and then determine the pitch/tilt angle of the top and bottom of the image. If you give me those I can crunh the numbers for you to work out the corresponding horizontal fov.