Forum

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Andrew

Pages: [1] 2 3 ... 6
1
General / Re: CPU and GPU benchmarks
« on: February 11, 2015, 02:32:36 PM »
Anandtech is using Photoscan to benchmark CPUs so you can pretty much find all your answers there.
Check out the recent 14core Haswell Xeons for instance: http://anandtech.com/show/8730/intel-haswellep-xeon-14-core-review-e52695-v3-and-e52697-v3/2

For the record, all of their Agisoft benchmarks were most probably recorded using pre 1.1 version, but I don't think there should be any game-changing differences between 1.0.x  and 1.1.x benchamrks.

-Andrew

2
General / Re: Workflow - Photoscan to game asset
« on: June 01, 2014, 02:57:29 PM »
Thanks for the kind words, and I couldn't agree more that for applications where immersion is key (hello VR!), photogrammetry is just amazing.

We are just starting a bit of a crunch to get our game finally out the door :) So I strongly encourage you to search the forum as it has a ton of useful information scattered. This might turn out more efficient than asking me - I might be slow and brief to reply in the coming weeks. Having said that, if you can't find what you are looking for, pm me and I'll see what I can do to help :)

3
Face and Body Scanning / Re: Mounting cameras to light stands
« on: May 14, 2014, 06:23:10 PM »
Anyone tried any of these? http://www.cullmann.de/en/results/cat/klemmen.html

I only know those tiny ball heads and they are rather not sturdy enough to keep Canon 600D's with kit zoom lens. They just might be okay for the light 100D with light prime lens though. The clamps themselves though might be adequate, while being half the price of Superclamp with stud.


4
General / Re: Aligning photos
« on: May 14, 2014, 04:56:48 PM »
Master your photography skills :) Use Aperture Priority on your camera with f8-f13 setting. Depending on amount of available light you might be forced to increase ISO. Remember to have enough overlap between photos that are supposed to align with one another (in case of project you uploaded - top ring of cameras were too high up compared to the other to rings of photos. Your top ring also looked almost straight down at center of scanned object so all these photos captured almost the same amount of information).

-Andrew

5
General / Re: Aligning photos
« on: May 14, 2014, 12:47:53 PM »
xiu2014, did you take a look at your photos in full resolution? Most of them have incredibly shallow depth of field, only a tiny portion of photographed object is sharp, the rest is blurry. No wonder Photoscan can't match these photos well. Also, top ring of photos is shot from too radically different perspective in relation to bottom rings. Photos within any given ring of photos align well to one another because each next photo is shot from only slightly different perspective (low angle between camera positions). But difference in perspective between any photo from top ring and any photo from other rings is just too significant.

Cheers,
Andrew

6
General / Re: SSD or graphic card?
« on: May 07, 2014, 04:00:07 PM »
8 gigs of ram is only okay for low-med quality reconstructions with limited amount of photos. Your GPU is fairly low end as well, but even with powerful GPU and just 8 gigs you'll still run out of memory on larger/higher quality projects and keep processing forever. When running out of memory, Windows tries to use a HDD swap file in place of memory which makes processing hundreds times slower. SSD is faster than HDD but that will only make your builds, say, 50 times slower instead of hundreds times slower, when compared to builds fit well within system memory. Definitely upgrade your RAM if you can.

There are some rough memory requirement estimates to be found on agisoft website, although I am not sure if they were updated once Photoscan switched to less memory-hungry algorithms...


-Andrew


7
General / Re: Recent work for games.
« on: May 06, 2014, 05:34:22 PM »
Nicely done, I have to reconsider processing at lower quality to get usable hair geometry :)

As for pant folds, these look awesome on static scan but can look weird when used on animated mesh, something to keep in mind. For our purposes, some folds worked out just fine, while others we had to get rid of.

-Andrew

8
General / Re: Recent work for games.
« on: May 05, 2014, 01:37:12 PM »
Nicely done Admir! Kudos for nailing minimal amount of photos to get great quality, most people would shoot ~3 times as many photos and probably end up with inferior results :)


Is that for some upcoming title or from Serious Sam 3?

-Andrew

9
General / Re: Single camera scanning thread?
« on: May 01, 2014, 09:08:40 PM »
I don't usually participate in this sort of discussion, and I will try to make it as brief as possible:

Please stop the nonsense of addressing Agisoft forum community as a whole in connection with some isolated 'hate email' that has NOTHING to do with this forum or this community. This forum has been nothing but helpful and respectful to anyone and everyone, even towards the strong opinioned and emotional users. I honestly can't recall a single post that would seem even remotely hateful, personal or otherwise inappropriate, including the original thread about single camera scanning that got deleted. It's only this here discussion that got unnecessarily heated. The beef anyone might have with one another, please keep us, the community, out of it.

Thank you,
Andrew

10
General / Re: Workflow - Photoscan to game asset
« on: April 24, 2014, 12:00:52 AM »
Autodesk wont be spitting out game ready assets. Yet. I believe that is their intention


Game assets are, and will always be, about efficiency as well as looks. For that reason I don't count on, I don't even want my photogrammetry software to build game ready assets for me. I will always be able to build assets that look the same but take, say, 15% less triangles and 15% better UV fill ratio and 15% crisper textures. And I'll always prefer to go the extra mile to gain few additional frames per second so the game runs smoother.

Having said that, I can certainly appreciate good tools that get me as far as possible with as little hassle as possible, but since functionality like smoothing out noisy mesh, building retopo or adaptive decimation already exists in other programs, I encourage Agisoft to keep improving speed and quality of photogrammetry-related algorithms first, and only when that is as good as it gets, to focus on further steps of asset pipeline. Actually increased precision, lower noise of reconstructions and better texture blending (that would pick sharpest photos for texture generation) would go a long way towards speeding up further steps in the pipeline.

-Andrew

11
General / Re: Workflow - Photoscan to game asset
« on: April 23, 2014, 11:22:21 PM »

If you don't mind my asking, from your PS scans, do you essentially process until you end up with an optimized mesh + diffuse map and a normal map? Do you use it to extract any other data? Would you suggest sticking with Zbrush for everything after PS, or would you suggest other tools?

For a lot of scans we find that noise is not much of an issue. Especially for rocks and such, we just use raw scans as base for further work. From there, we either decimate automatically and optimize further manually, or build retopo from scratch if we feel we can get polycount down considerably, or whenever that allows for easier and more efficient layout of UVs. Then we bake down normalmaps from highres, and occasionally we derive some specular maps as well, but that is tricky and far from ideal. As for specific tools we use - everyone has their favourite tools in the office, here's the list:
Zbrush (I actually hate the UI), Mudbox, Max, Topogun, Roadkill, Unfold3d, Balancer.

-Andrew

12
General / Re: Workflow - Photoscan to game asset
« on: April 23, 2014, 11:09:51 PM »
well for my taste, it looks too softy.. maybe more images would have done a better job though..


Did you mean this rock is too soft? https://www.dropbox.com/s/yw01c6ngy2vowwc/Rock_Ethan_Carter_closeup.jpg

It neither seems soft, nor does it need to be sharper - the whole cluster is no more than 1meter wide :) Perhaps you were referring to the smooth oval shapes? I'll pass on your complaint to Mother Nature :) Not being an ass here, just smiling at your comment as I actually often get this kind of feedback about the nature of what is being scanned, with my favorite being complaints about how bad/unrealistic/non-proportional some of our characters are. Good thing our actors don't read internet forums, they would hate to learn they supposedly do a really bad job of looking like proper human beings :)

Cheers,
Andrew

13
General / Re: Workflow - Photoscan to game asset
« on: April 23, 2014, 10:33:45 PM »
Gamegoof, I feel your pain with how your (previous) company approached photogrammetry :(

Having said that, I honestly believe Photoscan is not to blame here, what it does, it does wonderfully and last I checked (granted it was a while ago) it allowed A LOT more flexibility and a lot more tools for real artists than Autodesk solution. Surely Autodesk will not sit idle and will continue improving, but it won't spit out game-ready assets anytime soon, if ever.

You mentioned artists with no previous experience with photogrammetry cleanup, and with no previous experience with retopology/resculpting/baking, failed on their very first attempt. That is not surprising at all, and is hardly Photoscan's fault. I don't mean to be harsh, just trying to be fair with respect to Agisoft.

Let me break down main issues with photogrammetry-based pipeline, assuming person taking the photos did fantastic job and you have great highres source mesh:

- noise in mesh: unavoidable in real world, if you are going for very detailed mesh, you will get noise. It's not that hard to smooth out whatever needs smoothing, but surely requires some practice and, well, time. Alternatively, you can process scans at lower quality which produces much smoother results but retains a lot less detail.

- very high polycount and 'messy' triangulated mesh - I don't know of any software that automatically creates good topology and doesn't waste a single polygon. Some auto-retopology tools appear lately but all they do is generate 'pretty' regular topology, which may be of some use to animators, but for games you need adaptive decimation of your geometry (keep geometry only where you really need it, reduce polycount everywhere else), so that's really no good. I just don't see any alternative to manual optimization/retopology. No automatic solution will know where you need detail and where you don't.

- efficient UVs, this is something that Photoscan could improve, but there are so many UV solutions on the market that it's often more convenient to stick to one that you use on a daily basis.

- texture contains 'shading', by far the most problematic issue to overcome, unless you just want to keep geometry and create texture from scratch, or just don't care about in-game lighting/shaders. Few examples of this: any crack in the scanned rock contains 'shadow' in texture, and then lighting in engine puts shadow on top of that shadow, which results in unrealistic ubershadow :) If you get specularity or reflection, or Fresnel, in your scan-generated texture, it will appear static, while in reality those things change relative to viewing angle/light vector. All these things are just there in photos - this has very little to do with Photoscan. PS actually does try to reduce some of these issues (with enough photos from different angles, it will significantly reduce specular reflections on human face, etc.).

- geometry and texture is unique (non-tiling), which is great, but these assets eat up a lot of resources. I wouldn't recommend large-scale use of photogrammetry for games that need to run well/stream smoothly on previous gen consoles etc. Again, not something that Photoscan can or even should attempt to fix.

Games were always about a lot of work to create something that looks great and yet renders in fraction of a second. This doesn't change with photogrammetry - you need lots of experience and smarts in order to make use of scans, it's just that we collectively lack experience. But we'll get there! The really sad part is that big companies place so little value in research and learning of new tools and workflows...

-Andrew

14
General / Re: Workflow - Photoscan to game asset
« on: April 23, 2014, 12:05:07 PM »
(Excuse any typos, tbumbtyping on my mobile)
Photogrammetry output can only be as good as photos you input. They need to be pin sharp, with good dynamic range (well exposed, shot in raw, low ISO), without DOF issues, with good coverage and generous overlap, without obstacles between you and scanned object (think leaves or grass blades next to scanned rock),  scanned object needs enough surface detail definition for PS to see and match features, it cant bee too reflective, cant be transparent, ypu need to avoid light flare etc.etc.

In these ideal conditions PS will usually deliver results that are good enough 'out of the box' so you can use resultibg mesh as source 'hires' model. Even then though, you have to generate actual target mesh yourself. This doesnt really differ that much from typical game art pipeline,building separate hires sculpt an separate low poly geo to transfer normalmap and textures to.Sometimes you can get away with automatic mesh optimization to get low poly directly from highres PS mesh  (most 3d packages have tools for that, plus there are a few standalones like Antageo Balancer) but auto tools never give you as good results as manual retopo. Problem with decimating raw scan is, it contains noise. The more occlusion (worse photo coverage), the more noise you get in your mesh. Usually parts with most noise are of low importance and barely visible, but auromatic decimation will retain highest number of triangles exactly in those noise areas, wasting your polygon budget. If you are bent on using automatic decimation, you MUST smooth out most noisy areas first. Also, keep in mind that decimation built into PS is NOT for game assets, it is not adaptive and wont attempt to be smart about triangle distribution.
Now lets get real - you will rarely get photos perfect enough and so you will have to apply various amounts of manual labour to get that usable hires source mesh. This is the biggest problem when making studio-wide decissions about adopting PS in art pipeline - you never know how much work each scan will require. Sometimes you only need to create masks for (hundreds of) photos and do some mesh smoothing afterwards, sometimes you need to fight to even get proper photo alignment, sometimes you need to search for one photo among dozens that causes blurry texture, sometimes you need to resculpt broken areas, sometimes.... the list goes on :)

15
The one we used is http://p3d.in/ but there are other similar services, https://sketchfab.com/ and http://verold.com/ to name a few.

-Andrew

Pages: [1] 2 3 ... 6