Photogrammetric technology future of graphics?

Fire&Blood

Platinum Member
Jan 13, 2009
2,333
18
81
I'm never sure whether these graphics engine/tech topics belong here or in PC gaming forum but since this one doesn't necessarily exclude consoles, I'll post here and mods can move it if needed.

Wiki link: http://en.wikipedia.org/wiki/Photogrammetry

Apparently the costs involved are much lower than traditional approaches. It's great news if it can allow indie studios to close the gap in terms of graphics quality to the big budget studios:

Thorskan engine:
http://www.youtube.com/watch?v=oxLMOscTZRk

http://www.youtube.com/watch?v=6TEAtyUYZy8


The Vanishing of Ethan Carter:
http://www.theastronauts.com/2013/12/three-gifs-vanishing-ethan-carter/
 

Wall Street

Senior member
Mar 28, 2012
691
44
91
Such techniques usually produce voxel data. Voxels are interesting to study but generally considered slower to render than polygons except for certain limited situations. I would recommend reading about sparse voxel octrees if you want to learn more about the strengths and limitations of such methods. Voxel renderers tend to be slower than normal rasterization.

Scanning polygonal images is already a well accepted method. Look at the faces in any modern sports game.
 
Last edited:

Pottuvoi

Senior member
Apr 16, 2012
416
2
81
Actually photogrammetry usually creates point clouds, which are easy to convert into polygons. (and at least photoscan can create the polygonal textured version of the object.)

It's easy to create object with tens of millions of polygons and have very high resolution textures so moderation and proper work flow is needed to get decent asset for a game. (also decent amount of memory for computer, 18GB seems to be enough for ~20M polygons at the moment.)

Another big problem is that one has to be careful about getting correct colors without shadows, highlights, reflections etc.

Easiest way to use the method is to use it to get small assets or brushes for the more traditional workflow. (as metal gear solid 5 does.)

Thorskan goes and does the 'hard' way and they scan big areas as is, this needs very good photographer and a lot more beefy machines to create the environment. (also, they need to be careful on what kinds of environments to put into a game, trees, lakes and such do not work nicely..)

Here's a couple of 'raw' examples on what can be created quite easily. (decimated meshes and .PDF does require low resolution texture.)
https://www.dropbox.com/s/ir9mp05vahny86c/kyhmy_100k_4k.pdf
https://www.dropbox.com/s/8dgaapmdz86egnh/Aita_100k_4k.pdf
https://www.dropbox.com/s/5ckjqjksgedei01/Aitapala2_200k_4k.pdf
https://www.dropbox.com/s/bfgshbi4uwa5hzk/Betoni_20k_4k.pdf
https://www.dropbox.com/s/undax0dm1mz6lwt/seina_eps_100k_4k.pdf

These would need additional work to be able to use properly on game environment. (color correction, getting rid of lighting for albedo, displacement, normal, specular roughness, ambient occlusion maps etc.)
 
Last edited:

Fire&Blood

Platinum Member
Jan 13, 2009
2,333
18
81
Thank you both for shedding more light on the matter, pun intended lol.

I only have a minimal understanding of today's engines but even without it, it's just logical that it's very unlikely to leap ahead of the traditional, time consuming and very expensive methods without serious caveats. While hardware advancements open up doors for methods that weren't viable yesterday, major studios wouldn't be investing serious manpower and money in traditional methods as they do.

I did notice lack of shadows and a few other odd things. Also, lit animated objects/characters are missing and interaction with environment such as projectile impacts/collisions are yet to be shown off. Still, if it can help indies "cheat" and somewhat close the gap to industry leaders, I am all for it.

I remember an interview with Crytek's CEO early this year, he briefly addressed alternative methods. He dismissed them as being great for screenshots but stressed that Crytek focuses on improving their existing tech to get to photorealism that way. He also said that we are about 3 years out from doing in real time what rendering farms do today.