Photogrammetry

[Emor D'ni Lap]

I just finished playing the most beautifully detailed, naturalistic game I've seen so far: "The Vanishing of Ethan Carter".

I'm not saying it's the best game ever. It has some serious flaws in the gameplay department that I won't go into, and many people won't like the storyline much either.

BUT, this team knows their lighting, their animation, their texturing and UVing. They have studied nature: they know how a hillside slope is organized, how water flows and wind blows, they understand how things age and decay.

The game encompasses a LARGE volume of space. Not as large as, say, Oblivion's world. But far, far more detailed. This is a great demo of the Unreal game engine!

But the team's biggest tool in creating such realism is apparently the use of photogrammetry. I'd been just beginning to learn about it.
I'm not advocating its use in URU, though it may be possible to incorporate some of its advantages. In any case, it doesn't necessarily make things any easier, as this blog post from the "Ethan Carter" team indicates:
http://www.theastronauts.com/2014/03/vis...an-carter/
(Lots of great examples in this page - check it out!)

[Emor D'ni Lap]

After doing a bunch more research on the current state of cheap-or-free photogrammetry software options, it seems Autodesk's 123DCatch is still one of the best options. I have not tested Catch recently, but when it first came out I ran it through some sample images I made. Results were usable, but only after a lot of adjustments and manual patching. At that time, I felt that - if I had to build something to match a real-world object - I'd be better off sticking with normal manual polygonal modeling to image-plane references, in most cases.

The one really plausible alternative seems to be this guy's path: a combination of the Open-Sourced applications VisualSFM and MeshLab. He's written up a very comprehensive and detailed post on his whole workflow and thought process...it's worth reading.
And worth noting two of his points:
- VisualSFM is not easy to get installed, he says
- "123DCatch isn't that bad: ....it seems that Catch (Smart3DCapture) does a better job matching images and meshing them than my solution does some of the time, but that you have way more control over the input / output using my pipeline."

[Koller]

I have not tested Catch recently, but when it first came out I ran it through some sample images I made. Results were usable, but only after a lot of adjustments and manual patching. At that time, I felt that - if I had to build something to match a real-world object - I'd be better off sticking with normal manual polygonal modeling to image-plane references, in most cases.

Thanks Emor, I'm trying to use 123DCatch now. There's a lot to learn. How are you getting on with it?

[Emor D'ni Lap]

Thanks Emor, I'm trying to use 123DCatch now. There's a lot to learn. How are you getting on with it?

Hey Koller - to be honest, I had to abandon experiments with photogrammetry shortly after that last post, entirely for lack of time on my part.
However, that was almost two years ago, now! Since then, photogrammetery has come a looooong way.
And it's possible 123DCatch has evolved a lot, too. But there are now tons of other alternatives to explore, both free and at varying costs.

I worked with some folks last Fall who were using Capturing Reality, and they raved about how much better it was than anything else on the market....but it ain't cheap!
If you need to spend like $0.00, do a search for [photogrammetry + "open source"], or work through the free options in the Wikipedia list above.
The Regard3D software seems pretty well-regarded by reviewers, you might want to take a look at that.

Whatever direction you wind up working with, you'll also want to look into best methods for reducing the resultant poly counts for efficient gameplay, because *all* of these apps seem to crank out very-high-poly models (which are fine for non-realtime rendering, of course).

And please keep us posted on whatever you learn in your research!
I'll be getting back into some of this work later this year!