[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/3/ - 3DCG

View post   
File: 188 KB, 1920x1080, photo_scan_model.jpg [View same] [iqdb] [saucenao] [google]
712943 No.712943 [Reply] [Original]

Just how prevalent is photogrammetry in the industry currently? I've only worked in the industry for ~2 years and have noticed it show up in my work and job postings. Of course it's something that I was never taught in University, let alone a simple photography class, but it seems like a ton of studios are now using scanned models.

So what gives /3/? Should i put down my polygonal modeling and pick up a DSLR?

>> No.712945

>>712943

It's mostly used for super realistic stuff such as nature assets and food. I doubt photogrammetry will replace hard surface anytime soon.

>> No.712946

>>712943
If you're not artistically-inclined then snapping photos outside could be more your thing. Just remember that 3D scanning will never replace traditional 3D modeling/texturing in the industry though. It's just there to give the artists a really solid base to work on.

>> No.712954

>>712943
>Should i put down my polygonal modeling

Absolutely not. Continue modeling because there's only so much the real world can offer you as far as your imagination goes.

Example: if your art director wants you to model a piece of fruit that exists on planet Zoltron, you're gonna have to use your imagination to model that nonexistent fruit.

>> No.712957

>>712954
>he hasnt been to zoltron

typical standardlet

>> No.712966

>>712957
That's because he spent all of his money on standardlet software and now can't take a trip to Zoltron

>> No.712968

>>712946
>will never replace traditional 3D modeling/texturing in the industry
It absolutely will. It's already happening.
The "industry" doesn't give a shit about art and style, just churning out a product consistently and for that photogrammetry and 3d scanning is extremely low cost.
It won't kill poly modeling though because there will always be people who appreciate a different vision than just "mu realism".

>> No.712970

>>712968
>The "industry" doesn't give a shit about art and style

Of course the industry cares about art and style. That's why every game doesn't look like Modern Warfare or Halo.

Every game strives to be its own original content, from the initial idea to the last musical note as the credit ends.

If you've ever been involved in the creative process of making a video game, then you'll understand just how difficult it is to embark on something, you hope, people haven't seen before. It's a very tough business to succeed in.

>> No.712971

>>712943
Chief issue with photogrammetry is that you can't capture all kinds of materials. Have a translucent or metallic reflective object and bam you can't do it. Its a great tool to have in your belt for capturing certain kinds of materials and meshes. But it's not catch all, and will not replace "traditional" 3D production methods any time soon. That being said, having a working understanding of it and being able to work the benefits of reality capture into your own workflows is becoming crucial in the industry.

>> No.712972

>>712970
>Every game strives to be its own original content, from the initial idea to the last musical note as the credit ends.
What fucking industry are you talking about? The only ones who do this a indie developers. The INDUSTRY is ea, ubisoft, rockstar etc. They all make literally the same shit.

>> No.712973

>>712972
You're on the outside looking in, so I don't expect you to understand.

>> No.712974

>>712973
Good non argument.
You don't have to be "on the inside" to see the end result.

>> No.712975

>>712974
You're not a storyteller
You have no artistic skills
You're not a coder
Or a 3D generalist

All you do is spout nonsensical gibberish as to why everything looks the same without considering that the narrative is what drives the overall aesthetic of the game.

>> No.712977

>>712975
Someone's real upset.
Also wrong on all counts. Try again.

>> No.712978

>>712977
Go back to /b/.

>> No.712979

>>712978
Ironic.

>> No.712980

This place is filled with shit eating pseudo-theologians who doesn't have a single ounce of credibility backing them up.

>> No.712981

>>712980
Ironic given your clear lack of understanding for what the word theology implies.
Stay mad.

>> No.712989

This thread went to shit pretty quick. Good job, everyone.

>> No.713009

>>712945
It will the very moment AI-based synthesis is good enough. A decade tops.

>> No.713015
File: 149 KB, 476x476, 1565454898532.png [View same] [iqdb] [saucenao] [google]
713015

>>712966

No my boss spent part of his money on standardlet software, and knowing this software allowed me to easily find a job. Commie Open Sourcelets however can barely afford bread, let alone a trip to Zoltron.

>> No.713016

>>712980
get rekt nigger

>> No.713018

>>712972
have you been living under a rock or what?

>> No.713025

>>712972

Artwise? No they don't. Ubisoft has a totally different pipeline they only scan heads and even then they modify them heavily to make them more stylized. Rockstar does everything from scratch more often than not they don't even use photos in textures, EA are big into scanning athletes, the battlefield series and now Starwars. They scanned all the old movie sets and props.

>> No.713035

>>712943
In the future polygon based systems will be phased out and replaced with particle systems. Just a bunch of colored floating dots close enough together to form objects

>> No.713037

>>713015

based

>> No.713039

>>712943
So this is the power of textures.
Would it still be the same if that model was in quads?

>> No.713042

>>713035
That's retarded.

>> No.713043

>>713039
no if you removed the diagonal edges between the quads to model would be completely different <-sarcasm

>> No.713044

>>713043

I think he's talking about high poly raw triangulated scan data. Obviously the silhouette would be better because it's reading very flat despite the highly detailed texture. The problem with these things too is if you change the light source they stand out way too much.

>> No.713062

>>712968
>The "industry" doesn't give a shit about art and style,

Depends on the industry. I see film switching to photoscans. Animation and videogames I think are more likely to stick with their own stuff.

>> No.713203

>>712943
I don't know if it'll replace the 3D artists, but it's very useful in engineering fields. If you want to show a 3D model of anything with high exactitude, you want photogrammetry, not a beautiful representation (and innacurate)

t. a surveyor

>> No.713521

Good. Whatever makes the whinyass 3D-modeler princesses in the asset-production unemployed is fine by me.

>> No.713537

https://www.behance.net/yurijonk

>> No.713539

>>712943
Who do you think will be replaced first? The guy operating a camera or the guy who can create things that don't even exist to photograph in the first place.

>> No.713540

>>713042
Not retarded at all, there would be many advantages to that, but that is like the deep future when technology will be all startrek shit unless we've nuked ourselves to oblivion.

Not something that can be expected to occur within our lifetimes unless there is a massive breaktrough in computing and storage on the order of several magnitudes.

>> No.713546

>>713203
>a beautiful representation (and innacurate)
I assume you have specific engineering fields in mind and not all? Cause no photogrammetric methods currently come close to the pinpoint precision of a cad model for instance.

>> No.713584

>>713540
But what are the advantages of discrete points over interpolation along edges?

>> No.713648

>>712943
Is this some AAA exclusive thing or could I do something like that with a few pics with my DSLR and then get something decent?
How many pics do I need?
How long does it take to make something liek op's pic? (including retopo)

>> No.713649

>>713584
It'd completely remove the need for topology and UV's and allow us to focus on shape alone, like when creating surfaces made out of atoms in the real world.
But the most interesting things you could do with a continuous point-cloud model of things is the animation, deformation and surface to surface interface.

Models that can break apart anywhere act like fluids solids etc, fabric that can stretch rip or conform to any surface regardless of complexity.
Given infinite computing power that is the direction we'd be heading as that is analogous to how real matter behaves.

Conceptually voxels are really cool but you need physics simulation and resolution that is far beyond what we can do today
before they'd start overtake polygons in usefulness.

If one DeLoreaned to year 2200 I'd be surprised if technology would not be more similar to voxels than polygons.

>> No.713651

>>713649
>>713584

Like at some point there will be hardware that can do this type of thing casually in realtime.

https://www.youtube.com/watch?v=-5oDPCKUPPU

But looks grim that we'll be alive to see it given how we're fast approaching the end of 'Moore's law'. If we're unfortunate we'll only see consumer grade computers a few times more powerful than the ones we already have today rather than magnitudes.

>> No.713652

hey guys, real industry photogrammetry professional here. Literally was scanning clothes earlier today.

I got hired to do archival stuff, then went to a different company to get scanning into their substance and Houdini pipelines Now I am scanning for a vfx and design place in Los Angeles. Its really fun but quite tiring (long days shooting on set and sometimes working with live talent) and you are expected to know industry standard software. Not joking but blender does have some up and coming tools that look great- but that will never get implemented into most companies pipelines. Blender is a great content creation software and I hope more studios pick it up. But... EVERYTHING at least goes through Maya even if it's not created there. All pipeline tools are built in or around Maya.

You stupid fucks have been having this argument for 10 years. Learn both.


Anyway, I really love photogrammetry and its kinda my life's work and what I was born to do.

Its true it can't scan shiny surfaces, but there are other LIDAR scanners that can, and SLS scanners for extremely precise measurements. I just really really like seeing a irl object get turned into something we can share around the world digitally! For cultural archiving and artifact catalogues it's really great for museums and scientists. For games it's not the end all be all of game art- but it often can help artists get a better result faster and focus more on the overall art and design and not spending hours sculpting a small rock. It can even help artists create new assets by using Scan Data to mix into their own content. For VFX, it's HUGE in product spots. Shoes, clothes, people, fruits (scanning some mangoes for a commercial tomorrow), phones (to get precise measurements on the body to place into footage), Handbags, Sculptures, large set pieces, plants, and various textiles are all things I've scanned in the last 4 months. Its really interesting.

>> No.713654

If you don't mind a couple questions: Are you doing freelance? How did you get your first scanning job?

>> No.713732

>>713654
>If you don't mind a couple questions: Are you doing freelance? How did you get your first scanning job?

I am on contract with a company that does a lot of different things.
I went to gnomon and while they don't teach scanning there they teach everything else. The company reached out to me after seeing my reel at gnomon.

Other onset duties are important for a scanning artist such as making HDRIs of lighting, placing trackers on objects, and ensuring that directors /photography are shooting in proper formats.

Scanning is the bridge between what you see in the real world and what the artists can do with something in 3d. For example scanning a shirt irl that the houdini artist can then make "unravel" or for games scanning a selection of rocks that can be turned into a rock golem or something cool.
This will never kill the 3d artist, so don't worry about it.

We can even use scan data to train AI which I hope to focus on. Making a team of people to create a generative network to do style transfer on a 3d surface. (for example taking a stylized quick rock and creating a fully displaced huge model from it.) OR taking scan data and styletransferring onto it.

Would be pretty nice to greybox a level and hit go and get a quixel looking environment.

Pipe dream.. and probably not possible considering the huge dataset and computing power needed for it.

Also, you can start scanning with meshroom and a smartphone just to get the gist of it-no need to blow thousands on equipment till someone is willing to pay you for it.

>> No.713753

>>713652
I work with 3d scanners (structure light, lidar, photogrammetry) as an engineer for reverse engineering but I pretty much have free reign to do what I want with the equipment so I practise using it for making game assets.

One industry we have been trying to break into but having trouble is the game industry. Do you find that companies just prefer photogrammetry because it is easy to just capture alot at once? We use the Artec scanners so it's kinda a mix of photogrammetry (gets colors but based on structure light) and makes meshes no problem, it just isn't in the industry standard it seems.

>> No.713767

regarding game content, scans are fine refs, but that's all
you still gotta do all that shit by hand or it won't fit the pipeline

>> No.713788

>>713753
I think its more of the cost prohibitive factor desu, for the cost of one Arctec + software you could have 4 full camera setups + 4 realitycapture photoscan licenses and the people who know how to use them.

Also, I am super jealous! I wish I had more experience with SLS and reverse engineering. How did you get into it and does it require a degree in STEM?

>> No.713796

>>713788
Hmm interesting, I haven't looked to much at a cost for that setup but an Eva scanner with a license is ~23k so it does make sense that the cost is a larger investment at first. The benefit is it takes maybe an hour to process the data once you have it? With very minimal input besides some cleanup and alignments in the scan.

I have a degree in ME but to be honest you don't really need it for a scanner position. My boss has no college education and I learned it all while in the field, with maybe 1 day of training. Scanning with a hand scanner is a very underutilized industry, especially in the automotive. Hell we had to pass a 10k job recently since we couldn't make it to the area in the time they wanted (we work out of Oregon but the job was in Minnesota). I recommend looking at what scanning business are near you, or on the artec site (they have a database of resellers) and go from there. You'd be surprised how many are hiring for field engineers. Long as you know what to do to get a good result with photogrammetry it carrys over to scanning.

>> No.713813

>Art tries to capture reality for thousands of years, can't even get close
>Engineer builds a camera when he is bored and absolutely blows them all out of water
Artists really are a useless bunch, huh?

>> No.713814

>>713652
Nice dude.
I'm curious about photogrammetries implementation in things.

Like for example, the mangoes tomorrow, are they just gonnna be used as an asset in a commercial?
What did you mean when you said phones to get precise measurements on the body to place into footage?

>> No.713815

>>713652
I'm not a blendlet but I'm curious, you say everything goes through Maya eventually, I get that. But as far as just scanning and producing good models from photogrammetry, turning them into asset to be used elsewhere, how big is the gap in quality for the free softwares vs professional ones when it comes to doing photogrammetry?

Like, phone cameras are pretty good these days.

>> No.713816

>>713796
Is a field scanner just the equivalent of a pajeets retopo grunt?

>> No.713849

>>712989
Given that the OP was formatted identically to the standard shitpost seen on every ruined board, I think it's more a testament to how well this board has dodged most of that shit that the thread stayed clean of shitposting for as long as it did.

>>713025
That's interesting information, where do you learn things like that? I imagine their internal practice isn't talked that much, but I don't know much about it.

>> No.713876

>>713849

>That's interesting information, where do you learn things like that?

I watch the pixologic conferences. Other slides and testimonials from companies are posted time to time on polycount.

>> No.713983

>>712943
is that a fucking rock with raisins

>> No.713990

>>712975
Kek
You know an awful lot about this anonymous poster.

>considering that the narrative is what drives the overall aesthetic of the game

The only thing driving triple A executives is money. The lower level artists may sometimes have the freedom, time, energy and creativity to invest in the product but the industry at large loves their crunch, trend chasing and rehashing.

Overworking and heavily crunching just kills creativity.

>> No.714025

>>713816
Some days you scan rebar threads in the hot sun for 5 hours, somedays you scan ferraris and dinosaur bones. Since I am the only applications engineer I get to do sales demos as well which usually comes with a free lunch from the company I visit. Honestly a pretty sweet gig.

>> No.715195
File: 283 KB, 1600x1200, vertice.jpg [View same] [iqdb] [saucenao] [google]
715195

>>713546
Oh sure. It's good for the engineering fields that rely on measurements.
Cartography is 100% measuring, so it's pretty useful here. Also, it's way cheaper than 3D scanning
It's better to make measurements with surveying equipment and not just photogrammetry. A laser scannier is way better for construction than photogrammetry.
But photogrammetry is pretty good for creating otrophotos, where you don't need milimetric accuracy, but good enough one, like 1 metre or less.
Geodetic vertices can be used as reference points, because we know the exact coordinates

Pic related