3D Scanning Techniques and Workflows

[Infinite]

I thought I would start a thread here about scanning. Scanning on a small'ish budget.

A few days ago I was lucky enough to have into the studio Dawson James, a really talented actor, writter and aspiring film director. Dawson is a british actor and he recently played Greg Mutt, the youtube character created and animated by Paul Smith from Bustykelp.

**EDIT**

Really cool interview with Paul Smith who worked with Dawson on his virtual Dog navvi character, Greg Mutt:

http://www.cgheute.de/2010/05/interview-mit-greg-mutt/ (http://www.cgheute.de/2010/05/interview-mit-greg-mutt/)

http://www.youtube.com/watch?v=eQuToiab2uU

Dawson came in to be scanned for Paul's future animation and rigging work.

I used the 3D scanning equipment by DI3D (http://www.di3d.com/index.php) It takes stereo pairs of images to create range maps (depth maps) you can then use the photographs as texture information.

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson-01.jpg

Raw scan:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/scan-02.jpg

So I managed to get over 100 FACS scans of Dawson, featuring a range of normal expressions, emotions and isolated movements of his face changes. As well as some really high res texture reference using a Canon 5D Mk II camera.
http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson-02.jpg

Video Capture of the session:

http://www.youtube.com/watch?v=AaAnoGPIYR0

More to come...

www.ir-ltd.net (http://www.ir-ltd.net/home/)

triplegangers® (http://www.triplegangers.com/home/) (3D Scanning service and Digital Captures)

[Neox]

how small is small budget and which system did you use?

i have my own very cheap 3d scanning rig based on flexscan3d, but thanks to canon i cannot sync it for 2 cameras :/

1 cam worked pretty good so far for me, can post some of my results later

[Infinite]

how small is small budget and which system did you use?

i have my own very cheap 3d scanning rig based on flexscan3d, but thanks to canon i cannot sync it for 2 cameras :/

1 cam worked pretty good so far for me, can post some of my results later

Hi Neox,, I used a DI3D system and an Artec-M scanner.

When I say budget I mean in comparison to competing against larger VFX studio who have access to much more sophistacted equipment like, Cyberware PS systems or their own custom built setups. Most of these systems from the research I have done range between £50,000 to £250,000. Custom solutions can be much more.

I had allot less to play with. There is more information on prices available at the DI3D and Artec-Group website.

The Next Engine scanner is quite cheap and seems to produce fairly good results: http://www.youtube.com/watch?v=EblHQEdm-QM

I'm not sure why you can't sync your Canon cameras? that should be possible. I have used some software before from BreezseSystems that can synchronize upto 16 Canon cameras on one PC, they used it on the Matrix series. The Canon SDK is quite open and versatile as far as I know.

[Neox]

i only have vista or windows7 systems, so if i connect my 2 eos 400d to my pc they get the exact same ID in windows, which for the 3d scanner means, that there is only one cam, so if i shoot it shoots the same camera twice instead of 2 cams. The 1 cam system is a pain in the *** to calibrate. 2 cameras are supposed to be better but i only have eos cams in my surrounding and no matter which i tested its alway the same, windows gives them the same ID and the software can only reach one cam :S

my first scan:

http://www.polyphobia.de/nonpublic/jobs/2009/scans/3dscan_01.jpg


very fast calibrated and therefore pretty bad

gone through meshlab it has some potential

http://www.polyphobia.de/nonpublic/jobs/2009/scans/3dscan_01a.jpg

my second scan, still 1 cam so a lot of dead spaces

http://www.polyphobia.de/nonpublic/jobs/2009/scans/3dscan_02.jpg

i then closed the gaps with zbrush, and projection, the problem then are undercuts but all details stay, pretty decent it even has the cloth structure

http://www.polyphobia.de/nonpublic/jobs/2009/scans/3dscan_02a.jpg

however meshlab has some awesome gap filling tools which are really clever, never thought it would be so good, it loses detail but handles undercuts and overlaps pretty well

http://www.polyphobia.de/nonpublic/jobs/2009/scans/3dscan_02b.jpg

and my last test after a lot of testing, something i could actually use in a production back then

http://www.polyphobia.de/nonpublic/jobs/2009/scans/aermel.jpg

and the system itself is pretty damn cheap if you compare it to any other solution, however, i could never get 2 cameras to work due to the lack of nikons or other systems in my surrounding, so i got kinda lazy with my testing.
From all i've seen so far scanning is pretty much the same, 1 projectionsource (laser or beamer) and 1 or 2 cams campturing the progress, and this is what ich have too, i still hope to get my hands on another cam to test the 2 cam scan method, but well yeah, small hobby and no time, so it kinda sleeps:rolleyes:

http://www.polyphobia.de/public/pictures/Foddos/3dscan.jpg

the problem i saw with nextengines 3d scanner is the very small range it can scan, i planned on using my system for full body clothing scans and faces only on the side, and the system can definitely handle it, i just need to get 2 cams to work >_<

[Infinite]

i only have vista or windows7 systems, so if i connect my 2 eos 400d to my pc they get the exact same ID in windows, which for the 3d scanner means, that there is only one cam, so if i shoot it shoots the same camera twice instead of 2 cams. The 1 cam system is a pain in the *** to calibrate. 2 cameras are supposed to be better but i only have eos cams in my surrounding and no matter which i tested its alway the same, windows gives them the same ID and the software can only reach one cam :S

my first scan:


very fast calibrated and therefore pretty bad

gone through meshlab it has some potential

my second scan, still 1 cam so a lot of dead spaces

i then closed the gaps with zbrush, and projection, the problem then are undercuts but all details stay, pretty decent it even has the cloth structure

however meshlab has some awesome gap filling tools which are really clever, never thought it would be so good, it loses detail but handles undercuts and overlaps pretty well

and my last test after a lot of testing, something i could actually use in a production back then


and the system itself is pretty damn cheap if you compare it to any other solution, however, i could never get 2 cameras to work due to the lack of nikons or other systems in my surrounding, so i got kinda lazy with my testing.
From all i've seen so far scanning is pretty much the same, 1 projectionsource (laser or beamer) and 1 or 2 cams campturing the progress, and this is what ich have too, i still hope to get my hands on another cam to test the 2 cam scan method, but well yeah, small hobby and no time, so it kinda sleeps:rolleyes:


the problem i saw with nextengines 3d scanner is the very small range it can scan, i planned on using my system for full body clothing scans and faces only on the side, and the system can definitely handle it, i just need to get 2 cams to work >_<


Hi Neox, I remember those images when I was doing my own research into 3D scanning, they actually inspired me to look further, I also contemplated a flexscan system for a while.

Those scans are really cool, they pick up cloth well.

Perhaps what you need to do is try it on an XP or Vista platform? maybe think about a dual boot instalation to over come this?

Is Meshlab available to purchase standalone? I would love to see some videos of it in Action. The Artec Group software is really cool for aligning scans together and filling holes.

[Neox]

as said i have 2 400ds i would like to test with something else than canon :D

as for meshlab

its standalone and free to use

http://meshlab.sourceforge.net/

a bit crashy but still an awesome tool to combine scans or fill holes etc.

[Infinite]

as said i have 2 400ds i would like to test with something else than canon :D

as for meshlab

its standalone and free to use

http://meshlab.sourceforge.net/

a bit crashy but still an awesome tool to combine scans or fill holes etc.

Well I highly recommend that dual boot setup ;)

[Infinite]

Here is the first stage of processing.

Load in your generic base head/bust mesh, and the DI3D scan (including colour maps, we will use later) and also the full 360 grey scan from the Artec scanner.

1623

low poly generic control mesh

1620

artec scan

1621

DI3D scan

1622

You can use the DI3D scan as the true reference scale and align up the Artec group scan (this sadly doesnt support real world scale in the software) but both scans match exactly as they are so well callobrated on purchase.

Working with real world point information is invaluable and takes away most of the problems associated with using distorted reference images that contain fov or lens errors as well as any photoshop work that may have been done on the person.

The next stage is to go onto Shape Transfer using DI3D super cool landmark matching software package..

[Infinite]

The next stage is the Shape Transfer in DI3Dview.

Set out your landmarks in DI3Dview:

1627

You set out matching landmark points on both objects that correspond in position to each other. The software can then disort your mesh to match the other object. You have to be careful to keep the point numbering the same.

This technique is great on full body characters but you can end up with over 300 landmarks!

(This allows you to shape match two objects together that have different topology forms. I am amazed Zbrush doesn't have such an outstanding tool built in yet? I am sure you will see this type of tool every where soon. You can use this method on nearly anything!)

Select the areas of your grouped mesh that you want to conform or snap:

1628

press transfer!:

1629

[Infinite]

ZBrush Setup:

Then in ZBrush, go to your highest Subdiv level, in this case 6. Choose your layer you want to use as the character morph:

1630

Goto your lowest layer:

1631

Then import your exported shape transfered mesh from DI3Dview into your lowest level:

1633

(as you can see, this method keeps all your previous sculpture details from your last character project, which is quite handy for this new charaacter, although you will no doubt have to redo allot of the work but it means you can re-use things like eye lid structure, ears and other parts etc. Which all helps to bring the production time down from weeks, to a few days.)

and hey presto! your first stage of the projection process:

1632

(messy but a great blank canvas!)

Next stage, re-projection and cleanup...

[Infinite]

Projection Process:

First stage, using groups set out before hand. Mask off the areas you don't want touched. Make sure you StoreMT on Subdiv Level 06 before you start ; ):

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-project-01.jpg

ProjectAll all used on current object with scan visible in background:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-project-02.jpg

First ProjectAll on subdiv 02:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-project-03.jpg

2nd ProjectAll on subdiv 03:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-project-04.jpg

Clean up using Morph brush:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-project-05.jpg

Camparison of Scan and Lowpoly model:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-project-06.jpg

Animation switch, Project layer on/off:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/project-on-off.gif

Whole process so far? About 30 minutes. Some areas need work, like eyelids and ears, which will need matching to the hidden DI3D scan as well as allot of manual sculpting using reference images captured before hand.

Next stage, cleanup and detail sculpting...

[Infinite]

First stage sculpt.

http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-sculpt-01-sml.jpg (http://www.ir-ltd.net/uploads/cgfeedback/dawson/ZBC-sculpt-01.jpg)

Next step, texture baking from DI3D mesh to lowpolygon model in 3DSMax..

[PascalR]

Great thread, thanks for sharing Lee:thumbsup:

[Infinite]

Great thread, thanks for sharing Lee:thumbsup:

No problems, any chance you could change the title? sorry Pascal :o I put Wok! instead of Work!! :D

[PascalR]

Hehe sure thing , fixed :D

[Infinite]

I came across this today. Even if you have the money to buy a £100K scanner it doesn't always mean you can get good results http://www.youtube.com/watch?v=Pxb6UW9mSoY

What a waste..

What I would give to get my hands on a Cyberware PS system. Great technology but a bit pricey.

Hehe sure thing , fixed :D

Big thanks!

[Manuel Poehlau]

Great thread, keep on posting :thumbsup:

And nice to read some more infos about the Greg Mutt creators.
I was always wondering, who was responsible for this work.

[Infinite]

Great thread, keep on posting :thumbsup:

And nice to read some more infos about the Greg Mutt creators.
I was always wondering, who was responsible for this work.

Thanks Manuel, Paul Smith the guy behind the rig, has some amazing XSI rigging techniques. He should be sharing some information quite soon I think.

Base colour maps:

Next stage before actual baking of texture information from the scans, I re-use some old colour maps to act as a base.

1641

1642

[Infinite]

Clean up and new textures:

Here's a quick update of the new sculpt details and texture from the DI3D scan:

1643

I used Bodypaint 3D for the texturing, I can't wait for Mari ;) I will psot some screen shots of the baking process later.

Next stage, Mental Ray setup in 3DSMax2009...

[Manuel Poehlau]

Cool stuff!
And I really like the idea of re-using a generic base, which already has some basic scultping but also pore like details already. This can be a huge time safer, even if you resculpt things.

Those scan models are so cool for studying forms.
Most of the shapes and break ups are so subtle, i wouldn´t notice them in a photograph, and for me it´s the hardest part in realistic sculpting. I always tend to simplify and exaggerate forms too much.

[meshmasters]

Is Meshlab available to purchase standalone? I would love to see some videos of it in Action. The Artec Group software is really cool for aligning scans together and filling holes.

Meshlab is freeware. :-)

Cheers,

Joe

[meshmasters]

I'm not sure why you can't sync your Canon cameras? that should be possible.


It's totally possible and I've been doing it for years. I was an early adopter of 3D3 systems technology and I've been working with it for ages. You can certainly sync 2 canon cameras, I do it here all the time. The 3D3 system does take a LONG time to learn how to use properly. I spent at least a year of solid R&D really pushing the limits of the software and at times I wanted to kill myself... LOL. These days I can honestly tell you that I can get some of the highest quality scans possible out of that system, but you have to be willing put in a lot of learning time (sadly, many people give up too soon). I've also developed a lot of proprietary methods to make dealing with the hardware and software more user friendly. Still, I agree the calibration of the system is a pain.

Anyway, I assure you it is possible to link 2 canon cameras with the 3D3 systems software, provided that your particular Canon camera(s) are supported.

Don't give up!

Cheers,

Joe

[Infinite]

Hey Joe, great to see you posting on here. I think Neox was having issues because of his operating system. I haven't had issues with multiple Canons, wel l I used to with Vista as Canons SDK didnt support Vista until recently (about a year ago now)so I would imagine that could be the issue Neox is facing with Windows 7. My 9 Canons synch fine with XP and Vista (now) using Breeze Systems software.

Dimensional Imaging have only recently got access to update their software for their DI3D system to work with Vista 64bit.

Cool stuff!
And I really like the idea of re-using a generic base, which already has some basic scultping but also pore like details already. This can be a huge time safer, even if you resculpt things.

Those scan models are so cool for studying forms.
Most of the shapes and break ups are so subtle, i wouldn´t notice them in a photograph, and for me it´s the hardest part in realistic sculpting. I always tend to simplify and exaggerate forms too much.

Hey Manuel, yeah it's a real time saver and actualy makes the whole process gerat fun to do, as you can see your old work constantly improve, change form and adapt.

Here's another update, I just reused the shader from the last guy I did here that had darker skin and swapped the colour maps. This was the first batch of renders. It just shows how quick the process is, about 8 hours of work on and off.

1644

Next stage, fine details (eyebrows, skin fuzz, stubble and hair) and some other lighting setups to stress test the caucasian version of the shader...

[PascalR]

wow great workflow, your result is quite stunning for a 8h work:notWorthy:
Well done!

[Infinite]

wow great workflow, your result is quite stunning for a 8h work:notWorthy:
Well done!

Thanks Pascal, more to come.. The expressions are the fun part! :sw:

[meshmasters]

Hey Joe, great to see you posting on here. I think Neox was having issues because of his operating system. I haven't had issues with multiple Canons, wel l I used to with Vista as Canons SDK didnt support Vista until recently (about a year ago now)so I would imagine that could be the issue Neox is facing with Windows 7. My 9 Canons synch fine with XP and Vista (now) using Breeze Systems software.


Ahhh, I see. Yes, you are quite right, it could be due to driver incompatibility with the OS. I do have a 3D3 system set up at the moment that is running a stereo camera setup on Win7 64 bit. However, in this case I am not using the Canon's rather I'm using a stereo pair of uEye's machine vision video cameras. Come to think of it, I've not tried running a stereo pair of Canon's on Win7, so perhaps there is a compatibility issue. Actually, it would not surprise me if that were the case Canon does seem to take a long time to update their drivers. That said, overall I actually prefer using machine vision cameras with the 3D3 system, rather than using still cameras. Yes, still cameras will provide a higher quality image (in most cases) but the machine vision cameras are much faster, so it's a speed vs quality trade-off. If you are attempting to scan humans with a 3D3 system then speed will be your biggest issue. IMHO the 3D3 system is more suited to inanimate objects. Working with human subjects and the 3D3 system is somewhat difficult because of it's somewhat slow acquisition speed. That said, they have improved the speed a lot from what it was but I still think it's not the best choice for scanning humans. Where the system really shines is with still objects...you can obtain crazy HD quality, but the scan times will long.

Cheers,

Joe

[Infinite]

Ahhh, I see. Yes, you are quite right, it could be due to driver incompatibility with the OS. I do have a 3D3 system set up at the moment that is running a stereo camera setup on Win7 64 bit. However, in this case I am not using the Canon's rather I'm using a stereo pair of uEye's machine vision video cameras. Come to think of it, I've not tried running a stereo pair of Canon's on Win7, so perhaps there is a compatibility issue. Actually, it would not surprise me if that were the case Canon does seem to take a long time to update their drivers. That said, overall I actually prefer using machine vision cameras with the 3D3 system, rather than using still cameras. Yes, still cameras will provide a higher quality image (in most cases) but the machine vision cameras are much faster, so it's a speed vs quality trade-off. If you are attempting to scan humans with a 3D3 system then speed will be your biggest issue. IMHO the 3D3 system is more suited to inanimate objects. Working with human subjects and the 3D3 system is somewhat difficult because of it's somewhat slow acquisition speed. That said, they have improved the speed a lot from what it was but I still think it's not the best choice for scanning humans. Where the system really shines is with still objects...you can obtain crazy HD quality, but the scan times will long.

Cheers,

Joe

It would be great to see what these machine vision cameras can do, I belive the DI4D system might use them? but I am not sure. Also it would be great to get a glimpse of your 3D3 setup, asking allot I know :drool:

[Bigguns]

Hey Lee that really nice workflow! more complicated then I though it was but it's kinda fast and great awesome result.. in 8 hours! I would love to have a scenner too, it's really nice to learn the volumes even more.

Anyway, keep it up and do some Vray version too :) or send me the head and I do a vray test :)

[Infinite]

Thanks Bigguns, I will send over the mesh to you later. Just get me on MSN.

Another test update from yestarday. The flash shot, isn't so great.

1651

1652

1653

Expression tests to come today..

[Infinite]

Dawson's 'tripleganger' is getting closer to completion.

A quick test with hair, stubble and eyebrows before the expressions are setup. All made in Hair-Farm, over 1.5 million hairs, rendered in about 1 minute per frame:

1654

(I will post some screen shots of the hair creation process soon.)

Expressions next..

[Neox]

Anyway, I assure you it is possible to link 2 canon cameras with the 3D3 systems software, provided that your particular Canon camera(s) are supported.


i know it SHOULD be possible, though on my vista and windows 7 systems it is not i don't have xp licence floating around so i cannot test it on an xp system, the cameras are simply eos 400 d s and they are supported as far as i know :)
but it doesn't help me, on my machine i only have either one or the other camera ID in flexscan, depending on which cam i plugged in first and so the software recognizes 2 cams with the same ID and therefore only shoots twice with 1 cam as opposed to shoot one time per cam

ah and before i forget it again... very nice work, definitely makes me think to start my tests again :)

[Infinite]

i know it SHOULD be possible, though on my vista and windows 7 systems it is not i don't have xp licence floating around so i cannot test it on an xp system, the cameras are simply eos 400 d s and they are supported as far as i know :)
but it doesn't help me, on my machine i only have either one or the other camera ID in flexscan, depending on which cam i plugged in first and so the software recognizes 2 cams with the same ID and therefore only shoots twice with 1 cam as opposed to shoot one time per cam

ah and before i forget it again... very nice work, definitely makes me think to start my tests again :)

Hi Neox, perhaps it could be a bug of the Flexscan software? I wish I could help in someway, as that must be frustrating. Thanks for your comments.

Here's some screen grabs of one of the previous posts, baking the texture from DI3D scan to my low res mesh using Textureing Baking in 3DSMax. You can do the same thing in ZBrush but you are limited to the resolution of polypainting, which sadly doesn't cut it.

Texture baking settings. Most 3d apps have this I think.

1658

DI3D scan and low oply model ontop of each other. The lowpoly model has a displacement map applied, to match the high res scan. The DI3D model has the texture scan applied to it. This is what we need to extrapolate.

1655

Side by side and you can see the ray bake control mesh, which you can control.

1656

The final bake. It's dirty and messy but you can acquire some quite useful high resolution accurate information from it. As you can see it has been baked to my models UV layout.

1657

[Infinite]

3 shots with hair and peach fuzz layers:

1659

[Neox]

definitely a workflow worth trying out!
however i can't find any data about the pricing of any di3d software, is there any reseller around where i can get more information about their tools?

[meshmasters]

i know it SHOULD be possible, though on my vista and windows 7 systems it is not i don't have xp licence floating around so i cannot test it on an xp system, the cameras are simply eos 400 d s and they are supported as far as i know :)
but it doesn't help me, on my machine i only have either one or the other camera ID in flexscan, depending on which cam i plugged in first and so the software recognizes 2 cams with the same ID and therefore only shoots twice with 1 cam as opposed to shoot one time per cam

ah and before i forget it again... very nice work, definitely makes me think to start my tests again :)


Hello Neox,

Sorry to hear about all the trouble you are having with this. I don't doubt what you say is true and stuff like this can be so annoying.

Tell you what, as soon as I get a chance I will run some tests with FS3D running on my Win7 machine. I'll see if I can manage to get a stereo pair of Canon's to work properly or not.

I let you know what I learn.

Cheers,

Joe

[meshmasters]

Lee,

So I take it that the images that you are using to bake onto the mesh are the colour images from your DI3D rig? Yes?

I'm just guessing but given that the subject is sitting in the center of the frame are those images at about 45 either side of center?

If that's the case, would that not mean that you get a seam running down the center of the face when reprojected?

Just curious?

Cheers,

Joe

[Infinite]

definitely a workflow worth trying out!
However I can't find any data about the pricing of any di3d software, is there any reseller around where I can get more information about their tools?

Neox send me an email to admin "at" ir-ltd.net and I can get you some DI3D pricing information, or I would be able to at least give you the email information of someone there to talk to.

Lee,

So I take it that the images that you are using to bake onto the mesh are the colour images from your DI3D rig? Yes?

Joe


Hi Joe, yes correct.



I'm just guessing but given that the subject is sitting in the center of the frame are those images at about 45 either side of center?

If that's the case, would that not mean that you get a seam running down the center of the face when reprojected?

Just curious?

Cheers,

Joe

Yes and No. Well there is sometimes a faint seem. But as you can see from the captured data that has been baked, it's very scrappy anyway. You have to apply allot of manual clean up, hand painting and sculpting. It's a great base to work with though. I am in talks with the company about upgrading the system to feature cameras at the front and some at the sides, to over come this. It isn't really an issue though :)

[Infinite]

Time to pick and choose some expressions to work with. I think these might work well as some test material:

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson_008.jpg

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson_019.jpg

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson_069.jpg

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson_071.jpg

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson_072.jpg

http://www.ir-ltd.net/uploads/cgfeedback/dawson/Dawson_076.jpg

The next step is to workcareful with the layers in Zbrush to match to these expression scans.

Expressions next..

[meshmasters]

Hi Joe, yes correct.

Yes and No. Well there is sometimes a faint seem. But as you can see from the captured data that has been baked, it's very scrappy anyway. You have to apply allot of manual clean up, hand painting and sculpting. It's a great base to work with though. I am in talks with the company about upgrading the system to feature cameras at the front and some at the sides, to over come this. It isn't really an issue though :)

Ok, sure makes perfect sense. Just to be clear, I was not trying to imply that there were any problematic issues involved with this. As a matter of fact, this is exactly how my 3DMD scanner deals with mapping & projection and it's exactly how I work as well.

Cheers,

Joe

[Infinite]

Ok, sure makes perfect sense. Just to be clear, I was not trying to imply that there were any problematic issues involved with this. As a matter of fact, this is exactly how my 3DMD scanner deals with mapping & projection and it's exactly how I work as well.

Cheers,

Joe

Hey Joe, yes I know :) it was taken as so, I was just mentioning the fact that results need work. I can't wait for the day of a complete automated process but then we would be out of work!! :rofl:

Again I can't wait to see your pipeline and research results. I am really looking forward to sharing ideas with you in some capacity. :thumbsup: all your previous support, encouragement and ideas have really helped me move forward this year. Most appreciated!

[migusan76]

Great thread Lee! Thanks for sharing clap

[Infinite]

migusan76: No problem, I hope it helps in some way.

Thanks for the plug Pascal! :thumbsup: shame to see the other thread go :(

[PascalR]

No worries Lee, your thread was on my top row list anyways ;)

[Infinite]

First expression test,

Smile.

I am not happy with his Left side for some reason, I think it is the shape of the skull, I need to tweak it.

1666

More to come...

[Infinite]

and another, an awesome expression! Sorry Dawson! :rofl:

1667

[Intervain]

cool stuff Lee - an interesting read!

[collings]

great expression man, the second one is especially cool. The smile looks somehow a bit static not very natural if you know what i mean. but great nonetheless :)

[Infinite]

Intervain: I hope some of these points may be useful in the future :)

collings: Thanks. I know what you mean, the smile was actually quite staged so it didn't look natural in real life really, but the sculpt does need more work.

Last expression for the night. The inside of his mouth only has place holder textures, the gums are really unfinished.

Screech!:

1670

[tiktok]

Some great work here lee! very cool to see your workflow :sw:

-bayard

[madmatt]

I was missing this thread...

Great to see your workflow really interesting, there's a lot of work behind.

P.S.: thanx indeed to share with us who was behind my new myth "Greg Mutt" :D
I was trying to find some Infos about his creator without success...

[Kemono]

Wow, your really getting some awesome results in minimal time!
Do you paint any custom subdermal / specular maps, or you do you try to grab that from the color map that is scanned?

Fantastic thread!

[Infinite]

Kemono: Hi, no I try to keep everything to a minimum, well enough that I can get away with. I see many artists create multiple maps, loads of them, epi, sub, ss weight etc. When you don't really need them all that much and for speed it is just easier to use a few maps, but all characters and projects have different needs.

Currenty this guy uses Colour, Specular, Bump, Displacement, Gloss, Bacscatter Weight and that's it. The rest are controlled by procedural effects or colour slots.

tiktok: Thanks for commenting Bayard :)

madmatt: Thanks, at the start of the thread there is a link to how Greg Mutt was made by Bustykelp :thumbsup:

[piek]

Thanks for taking the time and trouble to compile this Lee, very generous and very informative.

[Infinite]

Thanks Piek :)

For anyone who may be interested, here is a direct comparison between MR and Vray. I still prefer MR's shader, as it is much faster than Vray and cleary has far greater Epidermis control where Vray is lacking in this area. It's pretty evident on the skin, as MR is able to eat in some of the light into the skins top layer, producing a lovely wax/skin effect.

This is a using an identical setup to Luc Begin's shader (although I am not able to wield it as well as Luc and the other Vray guru's here)

Mental Ray

1688

VRay

1689

Feel free to download the images, I recommend you flip between them in some kind of application. I use ACDSee 10. A superb image management program, has been since '98!

Any comments are most welcome.

[ZippZopp]

thanks for those comparison shots, really interesting to look at and analyze. you're right on that it looks like mental ray shader gives a little bit more control...are there attributes that the vray shader has over the MR one?

[MikeANash]

Looking sweet Lee.
Are you using Overall diffuse map with vrayfastsss2 ?
Looks like your diffuse is being washed away into your sss effects.
My setup for vraysss2
1692

[ZippZopp]

one quick question for you about your mental ray setup. are you comping reflection in or are you using the reflection settings built into the shader. every attempt i've done with the reflection on the SSS shader results in a really noisy reflection, even with a blurred hdr file.

[Bigguns]

That is not a good comparaison to me, Lee you had some setting not like mine, it's why it do not look better then the mr one and your sss is not strong enought it's why we feel it less.

I guess your next post with vray will be better ;)

I'm doing a convertion over mr of my character, it's sure it will be worse then my vray one ;l.

[Infinite]

Looking sweet Lee.
Are you using Overall diffuse map with vrayfastsss2 ?
Looks like your diffuse is being washed away into your sss effects.
My setup for vraysss2
1692

Hi Mike, thanks for posting. I managed to get Luc to help me out this one. Thanks for your tips though, I had my settings wrongly setup apparently. Vray is a beast!

thanks for those comparison shots, really interesting to look at and analyze. you're right on that it looks like mental ray shader gives a little bit more control...are there attributes that the vray shader has over the MR one?

Hey ZippZopp: I think Vray lacks allot with Epidermis control, as there are none but Luc, Manuel and Vimmy are the best to ask, they have got the best posted results so far.

one quick question for you about your mental ray setup. are you comping reflection in or are you using the reflection settings built into the shader. every attempt i've done with the reflection on the SSS shader results in a really noisy reflection, even with a blurred hdr file.

Hi, I always prefer to just render one image pass on characters. I know some artists render out dozens and dozens of layers then composite, even some VFX studios but I have yet to see great results from that process. I don't do anything special in MR but this is just for stills, I can almost guarentee if I was to render I would get fizzing and splotching happen.

Bigguns: Yeah thanks Luc.

[meshmasters]

What we need is a complete and comprehensive end-to-end vray tutorial for human character creation aimed at us mere mortals.

I'm really keen to learn more about rendering realistic characters with vray.

Any of your experts care to create one? I for one would be more than happy to purchase such a thing :thumbsup:

Cheers,

Joe

[Infinite]

What we need is a complete and comprehensive end-to-end vray tutorial for human character creation aimed at us mere mortals.

I'm really keen to learn more about rendering realistic characters with vray.

Any of your experts care to create one? I for one would be more than happy to purchase such a thing :thumbsup:

Cheers,

Joe

Hey Joe, I think Luc and Vimmy are the guys to push about that. I know part of the process but I am swarn to secrecy which I hate, as I share everything. Luc helped me allot but I don't think he wants to share the ideas, which is kind of understandable. Although I bet in the next 3-6 months tips and workflows will slowly leak out, as I see other artists here have learnt how to do it over time like Manuel and Mike Nash.

I would be more than happy to share my Skin settings with MR but those are already available on the models I sell.

[meshmasters]

Hey Joe, I think Luc and Vimmy are the guys to push about that. I know part of the process but I am swarn to secrecy which I hate, as I share everything. Luc helped me allot but I don't think he wants to share the ideas, which is kind of understandable. Although I bet in the next 3-6 months tips and workflows will slowly leak out, as I see other artists here have learnt how to do it over time like Manuel and Mike Nash.

I would be more than happy to share my Skin settings with MR but those are already available on the models I sell.


Understood but very disappointing nevertheless.

My own personal issue is that I know how to render realistic characters with mental ray... but I know nothing about about working in vray (although I'm very keen to learn). However, if no one sees fit to share it looks like this trend will continue on into the future - sigh :-/

As I said, most discouraging.

Oh well, I hate having to reinvent the wheel but then again, I'm capable of doing my own R&D. Besides, nothing in CG stays a "secret" for long anyway.

Back to the drawing board then.

Cheers,

Joe

[Infinite]

Understood but very disappointing nevertheless.

My own personal issue is that I know how to render realistic characters with mental ray... but I know nothing about about working in vray (although I'm very keen to learn). However, if no one sees fit to share it looks like this trend will continue on into the future - sigh :-/

As I said, most discouraging.

Oh well, I hate having to reinvent the wheel but then again, I'm capable of doing my own R&D. Besides, nothing in CG stays a "secret" for long anyway.

Back to the drawing board then.

Cheers,

Joe

As I say, normally I would share, as I share everything I can, always have but when something has been passed down from someone who has spent hours, days, weeks, months, testing and testing their own setup, which no-one had really done with Vray SSS2 before. I can't then just hand over Luc's settings without Luc's consent but perhaps if you contact Luc directly Joe, and work with him somehow :beerchug: I am sure he would be more than happy to help you. I have Luc's email, contact me offline and I can send it to you or you can get in touch with him here. He's a cool guy! :thumbsup:

I hope that helps. Honestly I think Vray is an utter pain in the arse to start with, once you get over a few hurdles it becomes more bareable. Still quite slow to render though IMHO.

..but as you can see from Luc's final results, it can look much better than anything else on the market, even any past renders from other artists dare I say it :sw:

[meshmasters]

As I say, normally I would share, as I share everything I can, always have but when something has been passed down from someone who has spent hours, days, weeks, months, testing and testing their own setup, which no-one had really done with Vray SSS2 before.

Hey, I totally understand.


I can't then just hand over the settings for nothing

Just to be clear, I was not asking you or anyone else to do that. I simply suggested that it would be nice if someone who is an expert at working with vray took the time to share this knowledge with others but if they want to hold this stuff back, fair enough, who am I to argue, I respect their decision.
But again, never was I suggesting that you should share trade secrets that have been obtained from others.


but perhaps if you contact Luc directly Joe, and work with him somehow :beerchug: I am sure he would be more than happy to help you. I have Luc's email, contact me offline and I can send it to you or you can get in touch with Luc here. He's a cool guy! :thumbsup:

Thanks, I appreciate that. Hey, I'm just like everyone else in the CG community who has been blown away by his mind blowing renderings and I'm keen to learn more. My point is that sure, I can put in the R&D myself, which it looks like I will have to do, fair dues.

I hope that helps. Honestly I think Vray is an utter pain in the arse to start with, once you get over a few hurdles it becomes more bareable. Still quite slow to render though IMHO.

I've been on the XSI beta team for quite sometime and I've never been able to get the kind of control with human skin that I can obtain with MR. I'm sure it's possible, I've just not been able to do it...yet. That's why when I see such superb renderings showing up on this forum and others it makes me want to take a second look and investigate this more deeply, which I will do. I was hoping to get a leg-up with this, but I totally respect peoples decisions not to share if they don't want to.

..but as you can see from Luc's final results, it can look much better than anything else on the market, even any past renders from other artists dare I say it :sw:

Hey man, no argument, I agree. IMHO Luc's recent work represents a major milestone in CG. The quality is off the chart and I have nothing but the utmost respect for his skill.

Cheers,

Joe

[Infinite]

No problem, all good.

I think the main benefit with Vray is it's interaction with sIBL and HDR rendering as well as the red scatter effect you get in teh shadow volume, something MR lacks in.

[Manuel Poehlau]

I really like the work on the expression, would be cool to see some animations with this guy :)

About the vray sss2. For me it seems to be much easier to get good results compared to MR. For some reason, I was never able to get good results with the MR FastSkin shader, it always looked too waxy. But I guess it was just me and I just didn´t find the right settings.

With vray, I´m pretty far away of the results from Luc or Vimmy. But it looks far better than anything I could do with MR.
And while it is always helpfull to get some tips/ideas from others, you only get the result you want to achive, by experimenting yourself.
I spend only about 2 weeks now with Vray´s sss2, so I think it shouldn´t be that hard to get some decent results.

Rendering skin is something that a lot of people want to learn, so it might be a good idea to make a thread where we can share our ideas, tips and tricks. What do you think?

[meshmasters]

Rendering skin is something that a lot of people want to learn, so it might be a good idea to make a thread where we can share our ideas, tips and tricks. What do you think?

I think that would be an awesome idea, let's do it! :thumbsup:

Cheers,

Joe

[Infinite]

I really like the work on the expression, would be cool to see some animations with this guy :)

About the vray sss2. For me it seems to be much easier to get good results compared to MR. For some reason, I was never able to get good results with the MR FastSkin shader, it always looked too waxy. But I guess it was just me and I just didn´t find the right settings.

With vray, I´m pretty far away of the results from Luc or Vimmy. But it looks far better than anything I could do with MR.
And while it is always helpfull to get some tips/ideas from others, you only get the result you want to achive, by experimenting yourself.
I spend only about 2 weeks now with Vray´s sss2, so I think it shouldn´t be that hard to get some decent results.

Rendering skin is something that a lot of people want to learn, so it might be a good idea to make a thread where we can share our ideas, tips and tricks. What do you think?

I know what you mean but trust me like others I have spent years doing just that, test, after test, after test. It's a long journey, while picking up tips along the way and I think its great to share but not everyone shares that philosophy sadly. I guess because we live in this stupid monetary system... if it wasn't for that and a certain other thing (which seems to control humanity) we would be off travelling the stars! :thumbsup:

I would be happy to share my Mental Ray shader results if that is of any help. Vray, that's up to you Luc and Vimmy and anyone else who feels they have mastered it :notWorthy: I would love to know how to use Vray correctly :thumbsup: there seems to be so many variables you have to get right.

Here are my results so far from Vray (thanks to Luc for helping me with this) they take bloody ages to render though!

Vray and sIBL setups (low quality settings, took 6 mins a frame at 800x600):

1695

I have to confess I still prefer Mental Ray and LW's Simpleskin. :dance:

I will post my Mental Ray shader settings later.

[Infinite]

Some extra shots, Luc's setup is pretty robust. I recommend everyone bombard him to open up his box of tricks :sw: I have lost count how many people on ZBC have requested for Luc's settings. 3 dozen? maybe more?

smartIBL lighting from the very talented and sharing Christian Bloch's site - http://www.hdrlabs.com/sibl/archive.html

1694

[Manuel Poehlau]

Yeah, looks better than the last tests !

Really wants me to do a realistic head too. But first I need to finish my stylized Ron Mueck based head :)

[Bigguns]

Hi there.

I'm sure if you do some test with the sss2 you will get really good result, spend some days on and you will see, really simple compared to the fastskin of mr.

[Manuel Poehlau]

I agree to what you said Luc. It´s much more fun to figure it out yourself. And on the other side, just copy and paste settings won´t give you the result from the guy who figured out those settings. They can vary very much, depending on the textures, the light setup etc. Maybe good for a starting point.

And also I think your great results on your latest head came from a great combination of models,bump and textures, the sss is just a small part of it.

:beerchug:

[Infinite]

I agree to what you said Luc. It´s much more fun to figure it out yourself. And on the other side, just copy and paste settings won´t give you the result from the guy who figured out those settings. They can vary very much, depending on the textures, the light setup etc. Maybe good for a starting point.

And also I think your great results on your latest head came from a great combination of models,bump and textures, the sss is just a small part of it.

:beerchug:

Its true to a point but its like the chicken and the egg, everyone takes from everyone else (they just don't always tell you that) The originators are the programmers who designed the software in the first place, these are the gods!

There are lots of pieces of information out there, take Max Edwins amazing work on Song Hye Kyo http://www.cgarena.com/freestuff/tutorials/max/songhyekyo/

Max was one of the first great artists to sacrifice and share his knowledge a very commendable thing to do. Not many people follow this philosophy sadly.

If it wasn't for people like Paul Debevec, James Cameron and the like sharing, we wouldn't be at the stage we are at this point in time, doing CG like this.

[Manuel Poehlau]

Sure thing. It´s always great when people share their knowledge.
If you´re in a position, that you don´t have a job, I think it´s reasonable
if you don´t want to give it away for free.
I´m also looking for a job right now, so I can understand where he is coming from.
On the other side, an indepth "making of" can go through all the big cg communities and can get you a job. That´s another away to think about that.

If someone shares its knowledge, it´s great, but still you have to figure out yourself, how to make good use of it. I think that´s what I wanted to tell :)


Its true to a point but its like the chicken and the egg, everyone takes from everyone else (they just don't always tell you that) The originators are the programmers who designed the software in the first place, these are the gods!

There are lots of pieces of information out there, take Max Edwins amazing work on Song Hye Kyo http://www.cgarena.com/freestuff/tutorials/max/songhyekyo/

Max was one of the first great artists to sacrifice and share his knowledge a very commendable thing to do. Not many people follow this philosophy sadly.

If it wasn't for people like Paul Debevec, James Cameron and the like sharing, we wouldn't be at the stage we are at this point in time, doing CG like this.

[Yolao]

Hi Lee.

Amazing work, is astonishing. Thanks for sharing your techniques.

Please, can you help me with this questions.

1-So, after you use DI3Dview to match your base poly head to the hi-res data you can also project the details from the scan and the colors right..?, but when doing an expression do you also use DI3Dview and then apply this as a morph in zbrush i assume, can you also project the hi-res wrinkle details from the scan into the morph....?

2-You have to match the teeth as well, do you also use DI3Dview for this..?

3-For a face rig, do you separate the parts of a expression shape in regions, like just the mouth, the nose, etc, and also in left right side?

Thanks, and congratulations to the founders of this site is wonderful.

Cheers

[Infinite]

If someone shares its knowledge, it´s great, but still you have to figure out yourself, how to make good use of it. I think that´s what I wanted to tell :)

Yes i totally agree with your points :)

Hi Lee.

Amazing work, is astonishing. Thanks for sharing your techniques.

Please, can you help me with this questions.

1-So, after you use DI3Dview to match your base poly head to the hi-res data you can also project the details from the scan and the colors right..?, but when doing an expression do you also use DI3Dview and then apply this as a morph in zbrush i assume, can you also project the hi-res wrinkle details from the scan into the morph....?

2-You have to match the teeth as well, do you also use DI3Dview for this..?

3-For a face rig, do you separate the parts of a expression shape in regions, like just the mouth, the nose, etc, and also in left right side?

Thanks, and congratulations to the founders of this site is wonderful.

Cheers

Hi Yolao, I will try and answer your questions. Thanks for posting here.

1- yes that's correct you can project those details in ZBrush. Although I prefer to use an external application for texture baking as you have to use Polypaint in ZBrush and you have to have a very high subdivided mesh, even at 8 million polys you don't get all the high resolution colour from the textures. Until Pixologic make texture painting in ZBrush resolution depended on the texture and not Polypaint it's not a viable option, well unless you have a super comuter and can work on billion polygon objects. But then again Mari should be out soon!

2- No the teeth are made by eye, you can find a free set on my models site ;) in the free section .

Although, Dawson is sending me his own set of teeth this week. He had moulds taken at the authodontist, so I will be scanning them with the Artec-Scanner for that extra realism.

3- Currently I do not rig, all these expression tests where to see if the pipeline was feasible, which it is. It only takes about 2 hours now per expression. They would be used as reference only for face rigs as moprhing face rigs alone is just to simple because they cannot simulate skin moving over muscle and bone. Great for reference but nothing more.

I hope that helps.

Here is another test from Lightwave. I have to say after using Lightwave for the last 15 years, I am now a Mental Ray and 3DSMax addict. It seems Lightwave is on the way out anyway now but I can still use it to tincker around with...... good bye old friend! :thumbsup:

Lightwave test:

1705

[meshmasters]

...Although, Dawson is sending me his own set of teeth this week. He had moulds taken at the authodontist, so I will be scanning them with the Artec-Scanner for that extra realism.


That is exactly how I work here. I make dental impressions and scan them at very high res. Then retopo, and insert in head.

This workflow provides for a very high degree of realism.

Cheers,

Joe

[Infinite]

That is exactly how I work here. I make dental impressions and scan them at very high res. Then retopo, and insert in head.

This workflow provides for a very high degree of realism.

Cheers,

Joe

It would be great to see some pictures Joe.

I will be posting my findings next week.

Cheers

Lee

[Yolao]

1- yes that's correct you can project those details in ZBrush. Although I prefer to use an external application for texture baking as you have to use Polypaint in ZBrush and you have to have a very high subdivided mesh, even at 8 million polys you don't get all the high resolution colour from the textures. Until Pixologic make texture painting in ZBrush resolution depended on the texture and not Polypaint it's not a viable option, well unless you have a super comuter and can work on billion polygon objects. But then again Mari should be out soon!



thanks Lee, what do you use for color texture projection.....bodypaint and 3dsmax i suppose?..

[Infinite]

thanks Lee, what do you use for color texture projection?..

Hi Yolao, if you check post #32 you will see that I used 3DSMax. :) but you can use any package that supports ray baking.

[Yolao]

Hi Yolao, if you check post #32 you will see that I used 3DSMax. :) but you can use any package that supports ray baking.

hey Lee, you answer before i edit my post..thanks...

[drummermenco]

it's great to see your workflow! great work as usual and thanks for sharing all this stuff!! :bouncy:

[msmbrush]

Infinite very very very informative Awesome Brake Down clapclap

[Infinite]

..thanks for your comments guys.

Here's another example of the quick turn around time from using 3D Scans, made this female test yestarday (she needs hair!)

The new girl:

http://www.ir-ltd.net/uploads/Emily-Big_Comp-01_sml.jpg (http://www.ir-ltd.net/uploads/Emily-Big_Comp-01.jpg)

Anyone have any crits on this, the eyes just don't seem quite right...

[meshmasters]

I don't know Lee, perhaps it's my monitor but often I find that the examples you produce using sibl look at bit strange to me, like there is too little contrast or something but there is something that bothers me about them. It feels like there is layer of white haze/fog being added. Perhaps it's because they are being displayed in isolation outside of the lighting environment. I'm sure I'm not doing a good job of explaining what I am seeing/feeling... but there is something that doesn't feel right to me. Maybe you'll understand what I'm on about.

Cheers,

Joe

[Infinite]

Joe, it might depend on how your monitor is set up?

I am using LWF when rendering out images, the renders check out fine on my calibrated monitors and look quite contrasted if I check on either of my laptops, so not sure what s going on there.

They all come out as 32 bit EXR images so I have allot of control over exposure and contrast, easy to fix.

Maybe anyone else might have some suggestions?

Cheers,

Lee

[meshmasters]

Joe, it might depend on how your monitor is set up?

I am using LWF when rendering out images, the renders check out fine on my calibrated monitors and look quite contrasted if I check on either of my laptops, so not sure what s going on there.

They all come out as 32 bit EXR images so I have allot of control over exposure and contrast, easy to fix.

Maybe anyone else might have some suggestions?

Cheers,

Lee

Hi Lee,

It could well be. The monitor that I use to view and communicate with on the internet is not calibrated in any special way. I'll try and have a look with some of my other machines and see if I notice any changes.

Cheers,

Joe

[Infinite]

Hi Lee,

It could well be. The monitor that I use to view and communicate with on the internet is not calibrated in any special way. I'll try and have a look with some of my other machines and see if I notice any changes.

Cheers,

Joe

Well your not the first to notice, last observation I had from someone was that my images were too dalk and contrasted! so it's hard to know who is using the right setup sometimes.

I just bought a Spyder Elite calbration kit, hopefully that might help me see what's going on this end, then I might know for sure which comments to talk on board.

[meshmasters]

Well your not the first to notice, last observation I had from someone was that my images were too dalk and contrasted! so it's hard to know who is using the right setup sometimes.

I just bought a Spyder Elite calbration kit, hopefully that might help me see what's going on this end, then I might know for sure which comments to talk on board.


I've got a Spyder too, and I find it very helpful. However, I think this issue brings to light a very interesting point (no pun intended...he..he..). That is, how much one can depend upon "proper" monitor calibration in the first place. What I mean is let's face it the vast majority of people viewing images on the web will most likely not be using any type of calibrated monitor. So as an a DCC creator what is one to do? Does one try and please the general public or does one try and please the technical community...or both?

My view on this is that it's a good idea to try and find the middle ground (if that is even possible). I try and view the images I make on a lot of different monitors, browsers, etc and see if any of them seem way off... if not, most likely they are good to go regardless of what the Spyder might say about it ;-)

Regardless, this is a difficult issue to overcome because unfortunately one can never be sure what kind of monitor the end user has or how it has been set up (or not).

In any case, I think this is a really interesting topic and I'd sure be interested in hearing other artists thoughts on this subject.

Lastly, Lee your new model looks really awesome! Great work as always man.

Cheers,

Joe

[Infinite]

Yeah I agree, and that's what I can't quite understand as the images look fine on my laptop, as my laptop generally displays images darker and more contrasted.

I know one of the lighting types in the picture might look a little washed out (2nd lot from bottom) but it doesn't look that bad to me. These are all also using a more widespread lighting setup from the sIBL sets.

Once I can figure out the ideal calibration (to suit the middle ground) I can easily tweak the renders in PS to match I think.

BTW your feedback is most appreciated, if people didn't comment, I would never know otherwise.

I will investigate further.

[meshmasters]

Also, just so you know, I don't think that there is anything major going on here. This is only a little niggle that I "feel" when viewing some of these...it's difficult to describe..there is just a little something that sticks out to my eye. This could be just me...I'm not sure, so don't drive yourself bonkers :-)

Cheers,

Joe

[madmatt]

Really nice tests Lee, I find the 3th orientation/set of iBLs the best because the others makes strange reflections on the eyes. For the problems of visualization I think that's because you render out with a black Background, if you could post with the hdri image in the back, we can compare the different feeligs.

Bye

P.S.: I use a Dell U2410 in sRGB mode and I see all well contrasted & lighted, maybe the highlights are a little bit bright in some sets, and not hyper contrasted but i found it good...

[Infinite]

Really nice tests Lee, I find the 3th orientation/set of iBLs the best because the others makes strange reflections on the eyes. For the problems of visualization I think that's because you render out with a black Background, if you could post with the hdri image in the back, we can compare the different feeligs.

Bye

P.S.: I use a Dell U2410 in sRGB mode and I see all well contrasted & lighted, maybe the highlights are a little bit bright in some sets, and not hyper contrasted but i found it good...

Hi madmatt, thanks for your comments and tips. What do you mean by 3th orientation/set of sIBLs?

[Manuel Poehlau]

I got really interested in this scanning workflow. While it is a little bit of cheating on one side, I think there is a lot you can learn while working with real world data. It´s not so different compared to traditional sculpters/maskmakers, who learn a good deal of information,when they work with face casts.
You can use this information for your own sculpts.
I think it´s much more valuable than working from distorted photographs.

I don´t have a 3d scanner, but I´ve got a really cool face scan from a friend.
It has a really nice asymmetrical face structure and I thought it would be cool to create a character/creature with this scan data.
There is no photo texture, just the mesh, so I will have to create the textures from scratch.

I am thinking of eventually making a step by step for my workflow,too. I will use some different tools and ways of doing it.So it might be interesting for some people, if I write down the problems and solutions I will have, while working on this.

[madmatt]

Hi madmatt, thanks for your comments and tips. What do you mean by 3th orientation/set of sIBLs?

I hope that's that HELPs, I allowed me to write some notes for explain you how I see the images in my monitor.

However I am convinced that if you put the head in a HDRI background we've diffrent feelings.

Bye

[Infinite]

I hope that's that HELPs, I allowed me to write some notes for explain you how I see the images in my monitor.

However I am convinced that if you put the head in a HDRI background we've diffrent feelings.

Bye

That's very helpful thanks. It's amazing you saw the reflection error on the 1st set of images, she was force using a a different reflection image in her eyes (which I removed for the later images) I will play around with the setup some more. Thanks for taking the time to critique.

I got really interested in this scanning workflow. While it is a little bit of cheating on one side, I think there is a lot you can learn while working with real world data. It´s not so different compared to traditional sculpters/maskmakers, who learn a good deal of information,when they work with face casts.
You can use this information for your own sculpts.
I think it´s much more valuable than working from distorted photographs.

I think I mentioned before. To me this is kind of like apples and oranges. It's like saying using a Wacom and Pen is cheating or that hand animation is much better than motion capture. It's just another tool to the trade and helps a great deal, removing most things that get lost in translation when creating characters from scratch. People that have the ability to do it well by hand, is fantastic. Some of us aren't that gifted though.

But your right, you learn so much more by working with the data, its a great way to understand the flow and structure of the human body. You learn so much along the way.

Although I have found with these scanners, they only really get you about 60-65% of the data you need, the rest is down to the artist to clean up the data and compensate by eye, let alone transferring the scanned detail and texture information over to a usable low polygon subdivision mesh, which is a whole other story and a tricky process to say the least.

Perhaps higher end, more expensive scanners are less work to process the data? I don't know.. :drool:


I don´t have a 3d scanner, but I´ve got a really cool face scan from a friend.
It has a really nice asymmetrical face structure and I thought it would be cool to create a character/creature with this scan data.
There is no photo texture, just the mesh, so I will have to create the textures from scratch.

I am thinking of eventually making a step by step for my workflow,too. I will use some different tools and ways of doing it.So it might be interesting for some people, if I write down the problems and solutions I will have, while working on this.

That sounds great, I can't wait to see. clap

[meshmasters]

Although I have found with these scanners, they only really get you about 60-65% of the data you need, the rest is down to the artist to clean up the data and compensate by eye, let alone transferring the scanned detail and texture information over to a usable low polygon subdivision mesh, which is a whole other story and a tricky process to say the least.

Perhaps higher end, more expensive scanners are less work to process the data? I don't know.. :drool:


Quite right...well said Lee.

Scanning takes you a good part of the way... but certainly not all of the way as you point out. No matter how expensive your scanner may be the data will need some amount of editing and that still requires patience, skill and knowledge. As any one who works with scan data on a daily basis will tell you, it's not the one button make art solution that many think it to be and artistic ability is still required.

IMHO the thing that makes scanning data particularly powerful is that you are working from datasets that have known values. It's not purely artistic guesswork and this can be an enormous aid setting up scenes, determining real world lighting values, or even trying to match objects accurately to a background plate.

Cheers,

Joe

[Rishi]

Infinite - Very informative and excellent workflow.I have some doubts about the material transfer option in DI3D,if possible could you please help me in that.
How does it work?I have done the shape transfer but not texture.

So it would be great if you describe the process for material transfer in DI3D.



Thanks

[Infinite]

Infinite - Very informative and excellent workflow.I have some doubts about the material transfer option in DI3D,if possible could you please help me in that.
How does it work?I have done the shape transfer but not texture.

So it would be great if you describe the process for material transfer in DI3D.



Thanks

I Rishi, have you purchased a DI3D sytem? Just curious.

I don't use the material transfer built into DI3D I bake the textures from one model to another in 3DSMax. See more info here on Page 4 - http://www.cgfeedback.com/cgfeedback/showthread.php?t=468&page=4

[majid_smiley]

Impressive !!! please share some render settings here !

[Rishi]

Infinite- Thank you for the info.No we haven't purchase it yet,we are trying on a demo version,if everything goes smoothly then we might buy it.

[Infinite]

Infinite- Thank you for the info.No we haven't purchase it yet,we are trying on a demo version,if everything goes smoothly then we might buy it.

Well I highly recommend it. Once you try it you will never look back, the hardware is scalable and the captured results are incredible. The shapetransfer software alone is worth it's weight in Gold. You can basicaly swap data between any meshes in minutes using matching landmarks. I'm amazed Mudbox or ZBrush don't have this feature. In fact no other 3rd party app has apart from Gator in Softimage but that is lacking.

It's also possible to create normal mapped data from captures to help create true bump mapping. Although a slightly more complex setup but possible.

http://www.ir-ltd.net/uploads/posts/plain-b_sml.jpg

http://www.ir-ltd.net/uploads/posts/roar-b_sml.jpg

[Levus]

It's incredible what are you doing Infinite! I'm very impressed!

And I have a question about meshlab that's Neox talking about.
Things I need is a meshlab software cameras and projector for start experiments with 3d scanning?
What I need to complete setup scanning system?

[Infinite]

It's incredible what are you doing Infinite! I'm very impressed!

And I have a question about meshlab that's Neox talking about.
Things I need is a meshlab software cameras and projector for start experiments with 3d scanning?
What I need to complete setup scanning system?

Thanks for your comments Levus : )

I'm not sure what you are asking here Levus? Neox, Meshlab? It's been a while since I posted on this thread.

To get you started you could look at www.scannerkiller.com it's a new affordable scanning service for artists. You might need some DSLR's to get you started. Canon 1000D's or 450D's. Or you could look at a system like www.di3d.com which is a bit more expensive.

[Levus]

I'm asking about http://meshlab.sourceforge.net/ software thats Neox using.
But it's seems it's only for editing scanned meshes, not for capturing.
I'm looking for some cheap or open source home-using software that I can use with DSLR's cameras.

To get you started you could look at www.scannerkiller.com it's a new affordable scanning service for artists. You might need some DSLR's to get you started. Canon 1000D's or 450D's. Or you could look at a system like www.di3d.com which is a bit more expensive.

Thanks Infinite, I will try www.scannerkiller.com for firts tests.

[Neox]

meshlab is used once you have your 3d data, for example to clean up the mesh, close gaps in the scan and such things, it has nothing to do with the scanning it self :)
you could also use it to crunch down the polycount of a highres sculpt from mudbox/zbrush i.e.

@infinite: affordable? i can't seem to find any information about any costs :O

[Infinite]

meshlab is used once you have your 3d data, for example to clean up the mesh, close gaps in the scan and such things, it has nothing to do with the scanning it self :)
you could also use it to crunch down the polycount of a highres sculpt from mudbox/zbrush i.e.

@infinite: affordable? i can't seem to find any information about any costs :O

If you sign up and dig around, there is some pricing information available. It is cheap.

Levus: Apart from ScannerKiller, I don't know of any other cheap Stereo Photogrammetry software.

[eRKK]

Hey! I've admired your work for a long time now and it makes me happy to see that you say you'll gladly share your MR-skin settings! I've got a head model going on for my future demo-reel, hopefully I'll be able to get a WIP thread up one this newfound amazing forum. I haven't worked much whith skin shaders except for in Cinema 4d, and needless to say it was almost impossible. Since your skin shaders are some of the very best I've ever seen I'd love to know of your techniques. Also, to learn some important "notes" on skin. Such as the importance of a good bumpmap, reflections, variation in skin color and specularity etc.

I believe that I've done my fare share of searching for good explanations on skin shading, and I'm amazed to find that there seams to be no books on it whatsoever. Oh, how I'm ranting on... Anyways, anything that you'd like to share would be much much appreciated! :beerchug:

[Infinite]

Hey! I've admired your work for a long time now and it makes me happy to see that you say you'll gladly share your MR-skin settings! I've got a head model going on for my future demo-reel, hopefully I'll be able to get a WIP thread up one this newfound amazing forum. I haven't worked much whith skin shaders except for in Cinema 4d, and needless to say it was almost impossible. Since your skin shaders are some of the very best I've ever seen I'd love to know of your techniques. Also, to learn some important "notes" on skin. Such as the importance of a good bumpmap, reflections, variation in skin color and specularity etc.

I believe that I've done my fare share of searching for good explanations on skin shading, and I'm amazed to find that there seams to be no books on it whatsoever. Oh, how I'm ranting on... Anyways, anything that you'd like to share would be much much appreciated! :beerchug:

Hi Erik, Thanks for your nice comments.

I don't really use MR or Vray now but there are plenty settings floating around the net and around these forums that might help.

Lee

[Ed3d]

Hi Lee, really great pipeline you got there.
I think DI3D's system is pretty cool , I've been checking it out lately.
I know it's kind of an old thread but how did you manage to convert/change these 4 regular photos to contain normal map data.

Thanks

[Infinite]

Hi Lee, really great pipeline you got there.
I think DI3D's system is pretty cool , I've been checking it out lately.
I know it's kind of an old thread but how did you manage to convert/change these 4 regular photos to contain normal map data.

Thanks

Hey Elad,

Yeah they are good. I reverse engineered some USC ICT Papers on Normal Map generation. You can acquire normal maps from captures by using a trigger box for the DI3D kit (which costs about £3k) that fires multiple lighting configurations in the space of about 10 seconds. The Canon Cameras can buffer up to about 10 images each.

From those 6 lighting captures - top, bottom, left, right, front and back you can do math on the images to produce the data you need. You then can combine the images together using the RGB channels to get the very rough normal maps.

A clue is XYZ = RGB :sw:

but it's all useless unless you have the code to convert the Camera Space Normal Map data to deform the geometry. As the Camera Space Normal Maps are of no use at render time in 3D space because they are set at a fixed point from the cameras capture position. This they don't tell you about in the papers. As the code costs allot of money to license or use directly. There is nothing commercially available yet, as far as I know.

[Ed3d]

Hi Lee

Thanks for your quick reply.That's a nice trick , but from what I understand at the end of this process the data cannot be used on the model since the code is expansive? did you tried to convert those normal map photos into a NM texture and apply it on the mesh?
It's kinda hard to hold a facial expression for 10 seconds , especially if its extreme which makes the data less accurate , but never the less it still looks like it can produce much needed extra detail when doing face scans.

[Infinite]

Hi Lee

Thanks for your quick reply.That's a nice trick , but from what I understand at the end of this process the data cannot be used on the model since the code is expansive? did you tried to convert those normal map photos into a NM texture and apply it on the mesh?
It's kinda hard to hold a facial expression for 10 seconds , especially if its extreme which makes the data less accurate , but never the less it still looks like it can produce much needed extra detail when doing face scans.

You can't convert the Camera Space Normal map data into anything else usable, it doesn't work like that. Plus you can't apply Normal maps to a mesh, you can extract of course, World Space or Tangent from a previous model in ZBrush or another application. Camera Space Normal maps are no good in 3D space if your rendering camera moves.

What you need is the code to deform the geometry of the mesh using the Camera Space Normal map, not just at render time.

There are also other algorithms that can be used to Align the images before you generate the camera space normal map texture but again they cost and it's tricky getting access to them.

The trick is getting the external code : ) until someone releases something commercially available. Which wont happen any time soon, as some very big companies own them and their patent.

I hope that helps.

Lee

[Cutia]

Hey guys, nice discussion happening here.

I think you should take a look at Photoscan, a software made by Agisoft
www.agisoft.ru

It's a photo based scan, and cost around 100 euros or so.

You only need one camera that records the equivalent lens on a 35mm film on the exif data of the image.
Any commom DSRL should do it, and the software give some impressive results juding by the price.

here is a test that I've maked of a friend.
I've taked the pictures at my room, only using the lightning comming of my window, next week we will try make a studio session with some pro lights.

take a look at the results.

http://vimeo.com/21225495

See ya.

[Infinite]

They look like incredible results, thanks for sharing Cutia. Investigating now. :bouncy:

[piek]

I agree, Agisoft have some very nice kit. A few weeks ago I tried their free StereoScan with a consumer 3D camera/MPO files and got great results. Certainly made me want to upgrade to PhotoScan.

[mala]

This looking likes a very interesting bit of software,especially for the price!

@Cesar..how many photos did you take for the model you have in video posted?.....and did you move around the subject or did you rotate the subject whilst taking photos?

I've been having a play with mac version trial,and even with some pro very hi res pics i'm having trouble getting it to align more than 2 photos most of the time,which of course means it fails to build the model.

the pics i'm using were all made by rotating the subject and i'm wondering if that makes a big difference.

Cheers,
mala

[Cutia]

Hey guys, the software is very good indeed, I recommend it to everybody.

@mala
Static object, and I rotated around it.
I'm sure that it makes a great difference, cause the software usethe intere photo to align.

I've taken 21 photos. and they all aligned perfectly.

You should try with static object, im sure it will work.

See ya guys

[Infinite]

This is the business:

http://www.youtube.com/watch?v=JX5stsU6xfE

Some how I don't think this will be publicly available for a good few years yet.

High-Quality Single-Shot Capture of Facial Geometry

Their custom Highpass filter is incredible.

[madmatt]

Ooh great news!!!

Really interesting, the future seems to be very interesting :D

[Vimmy]

just 2 test i did with agisoft,

Quite cool what you can get with few simple photo.

around 10 pic, so i guess with more, you can get probably nicer results... and higher setup for computation to.

http://dl.dropbox.com/u/14985511/CGFEEDBACK/quicko_scano.jpg

[Infinite]

Awesome tests Vimmy!! this is the sort of stuff they need on their site to promote it. Did you find it easy to use?

I wonder if anyone has tested using a structured light pattern? it seems to love skin.

[Vimmy]

Yea it totally friendly to use. very very intuitive. :)

I only test with some basic outdoor shots with a 7d. but yea i would like to test with pattern, i'm sure the result could be much nicer.

[PascalR]

Very cool test Vimmy :bouncy: will have to give a go at some point, nice!

[Infinite]

Yea it totally friendly to use. very very intuitive. :)

I only test with some basic outdoor shots with a 7d. but yea i would like to test with pattern, i'm sure the result could be much nicer.

Cool. Is it of you? :)

[meshmasters]

Wow, great job...very impressive!

Cheers,

Joe

[Vimmy]

Cool. Is it of you? :)

Hehe no, only the shoes, not the face;)

[Infinite]

Hehe no, only the shoes, not the face;)

ahh shame. I thought we might be starting a digital clone renaissance!!

[Vimmy]

:)




The message you have entered is too short. Please lengthen your message to at least 10 characters.

[moonjam]

Holy crap - these are amazing!

Definitely going to give Agisoft a go. clap

[Vimmy]

AJ::o yes you should give a try.



just a little point cloud, with a new photo set, took around 100 shots for that one. i dont know how the soft will deal with it, but, i hope that I could get better details. Just for comparison i post the point cloud of the last head test.

anyways i'll post details soon of those tests. And maybe some from hd video to. (extrated pic)
http://dl.dropbox.com/u/14985511/CGFEEDBACK/com.jpg

[Infinite]

Awesome, thanks for sharing Vimmy! Looking forward to seeing more from you :drool:

[oglu]

looks really cool... but how does the tool know where to put your image...?
i mean all those blue cards around... have you placed them by hand..?

[piek]

I'm guessing* it's 'scale-invariant features' and 'bundle adjustment'? Judging by the need for images that overlap.

http://en.wikipedia.org/wiki/Scale-invariant_feature_transform

http://en.wikipedia.org/wiki/Bundle_adjustment


Microsoft's Photosynth does something similar and there are a number of toolkits that allow you to convert your synth to 3d. Agrisoft's program is better as the whole process can be done offline, you don't have to upload your photos to a server.

*Let me stress guessing

[ulf]

really cool, thanks for sharing vimmy! gonna check that out!

[mala]

@cutia..thanks for the tip that's helped a lot,getting better results on aligning the pics now.

@vimmy..the first mesh you posted,did you do any clean up on that in photoscan or other soft? That mesh looks amazing considering the pointcloud does not look very dense.What build mode/settings did you use?

I'm really impressed with your 2nd pointcloud, there's a lot in there.I did a couple of tests yesterday with some pics taken with a 450D..so a bit less resolution than you have,but it can't find anywhere near that amount of points on the skin.
Build times have been very long for me on a 3.2, 8 core mac(10GB Ram),i only have a 8800GT card which i think is right down the bottom of what photoscan will work with. Photoscan also froze up on me during one build,so guess i'll have to look at cutting the mesh up into chunks.

@Infinte..i googled 'structured light pattern' to try and get my head round that..i have access to video projectors and a myriad of intelligent LED lighting due to my RL job.
So i'm interested in making some tests..one thing i don't really get though is how you could still retrieve a correct texture if you take pics using structured light patterns.

More shooting and tests tomorrow i think!
Cheers,
mala

[mala]

Here's some tests i did this afternoon,just pointclouds no Geo build yet..

Weird thing happened with the cloud for the Mannequin..it built most of the room as well!...so you can play join the dots and guess whats in the room:p

There are 2 mirrors in the room which Photoscan partially modelled the reflected objects in as well..thought i was having a Blade Runner 3D photo episode for a moment.

will hopefully grab some people at work tomorrow and try to get some good faces.

Cheers,
mala

[gjpetch]

This thread inspired me to buy agisoft photoscan and give it a go myself. My first attempt, 28 photos taken with a crappy point-and-shoot canon ixus 800is, and some bump details (just a high pass filter on the photo, got the idea from the stuff Infinite has been posting, can't believe it works!)

[Vimmy]

@vimmy..the first mesh you posted,did you do any clean up on that in photoscan or other soft? That mesh looks amazing considering the pointcloud does not look very dense.What build mode/settings did you use?




it was straight out of the Agisoft. and use high with smooth for extraction. and yea i let it run during the night... cause it probably took over couples hours.


i also generate a mesh from the other more dense point cloud, and suprisingly result was similar to the first one... but with medium setup... and i reduce my photo to 75 or so, cause it crash with 100 18 mp pic...




gjpetch: Quite nice with that high pass

[mala]

Here's my quick results from some photos of the guys at work.
http://img850.imageshack.us/img850/3272/98883446.jpg

http://img202.imageshack.us/img202/6316/13650169.jpg

http://img29.imageshack.us/img29/7826/52754197.jpg

Took the pics outside in the shade and made the mistake of having a black painted brick wall behind them,which has merged into the hair as well.

About 50-60 photos,Medium+Smooth Build settings and 3k map...opened in ZB with SkinShade4 on just for quick test.Of course they would need a lot of clean up...and i didn't take any pics of the top of their heads..(they are all taller than me;) )..so some very weird hairstyles have appeared.

I quickly removed some big chunks of 'smoothed' geometry in Photoscan,hence the holes.

Loving Photoscan,now i have to learn to fine tune it to get more detail and work with it all in ZB.
Cheers,
mala

[Vimmy]

Second head is quite cool.:). did you try less photos but with high rather than medium for extraction? I got better result that way.

[Infinite]

Mala & Greg, just incredible results. This soft is very powerful indeed for price. Super interested to see more results. :beerchug:

[mala]

Vimmy,just running a test now with fewer Photos and higher settings,will post results when its done....may be a while.

@Infinite..it sure is cool!...i couldn't believe the results i was getting to start with and only on medium settings.

Having looked back over my photos,i think there is quite a bit i can do to improve the images i use in photoscan to hopefully get better results with more detail and quicker.

So i'm making a plan to take some photos indoors with a few different light setups to see which works best.

I noticed on my photos for the guys above that the lighting had changed a bit between some shots(sun finally came out from clouds a rare site in London at this time of the year:) )....this caused some of the pics to be over exposed or have 'hot spots' which i think is what created the lumpy bits on the nose of the second head.
Also some of the photos have areas that are quite badly focused/blurred which of course is going to lead to less detail.Got to learn to use my camera better!

With the kit i have at work,i can quite easily build a large diffused 'dome' of light around the subject or various other setups..to get even controllable light,but with photoscan using the background in the photos to align the images i'm not sure how well that will work out...will have to think about that one.
Thinking about building some kind of slightly heath robinson rig/camera jib so that i can rotate around the subject quickly and yet have the camera steady ...perhaps even with some kind of vague calibration like 10 degree steps..or whatever the bare minimum needed for overlap of the photos is.

Is it obvious i'm addicted already? :rofl:

Cheers,
mala

Nice results by the way Greg....i'm gonna have to look into using the pics to add detail as well i think.

[Infinite]

Vimmy,just running a test now with fewer Photos and higher settings,will post results when its done....may be a while.

@Infinite..it sure is cool!...i couldn't believe the results i was getting to start with and only on medium settings.

Having looked back over my photos,i think there is quite a bit i can do to improve the images i use in photoscan to hopefully get better results with more detail and quicker.

So i'm making a plan to take some photos indoors with a few different light setups to see which works best.

I noticed on my photos for the guys above that the lighting had changed a bit between some shots(sun finally came out from clouds a rare site in London at this time of the year:) )....this caused some of the pics to be over exposed or have 'hot spots' which i think is what created the lumpy bits on the nose of the second head.
Also some of the photos have areas that are quite badly focused/blurred which of course is going to lead to less detail.Got to learn to use my camera better!

With the kit i have at work,i can quite easily build a large diffused 'dome' of light around the subject or various other setups..to get even controllable light,but with photoscan using the background in the photos to align the images i'm not sure how well that will work out...will have to think about that one.
Thinking about building some kind of slightly heath robinson rig/camera jib so that i can rotate around the subject quickly and yet have the camera steady ...perhaps even with some kind of vague calibration like 10 degree steps..or whatever the bare minimum needed for overlap of the photos is.

Is it obvious i'm addicted already? :rofl:

Cheers,
mala

Nice results by the way Greg....i'm gonna have to look into using the pics to add detail as well i think.

Awesome idea. Just out of interest where do you work?

The most important thing with soft like this is pattern and detail! So sharpness and clarity is a must. Try to minimize DOF and motion blur as much as possible.

I want to test out this software at some point. I have about 10 DSLR's laying around if anyone wants to buy em! : ) Canon 450D's 12 MP

..which can be synced and fired using http://www.breezesys.com/ (developed for the Matrix bullet time shots I think)

I friend (Alexander) ran some Scanner-Killer full body captures through the soft.

http://www.ir-ltd.net/uploads/results_01.jpg

http://www.ir-ltd.net/uploads/results_03.jpg

Scanner-Killer process:

http://www.ir-ltd.net/uploads/sk_fb-02.jpg

These where captured with carefully calibrated twin Canon5D MKII's. You can see the other soft does a great job but they lack a little depth.

[mala]

Hi Lee,
I'm a freelance lighting tech/rigger,so i work all over,but atm i do a lot of work for a company based in Shepherds Bush.Over 10 years in the biz,means i can get me hands on all sorts of lighting+video kit..

Yep, trying to minimise DOF and motion blur is what makes me think a some kind of rig would help...a tripod is ok until you need to move round the subject quickly.
I took some photos last year with same camera(450D) just to use as ref and for skin textures(no use for photoscan 'cos i rotated the subject and the soft can't align them well)..but that time i used canons control software that comes with camera,to preview the sharpness of the focus and trigger the camera from laptop.i might try the same this time,but will need a long usb lead and some way to stop it getting tangled round the rig.!

My early tests with the mannequin and a friend who happened to be wearing a patterned top showed me how much patterns helped the align process..maybe i need to print a t shirt with a crazy test pattern on it and make all the subjects wear it:)

I've been reading up on scanner killer,the results look great...but for the moment i'm going for the cheap option of just one camera and photoscan.....although the idea of those 10 450D's all synced up sounds tempting too!

Cheers,
mala

ps those scanner killer pics look like you made a bad deal with Jabba the Hutt and got frozen :-)

[mala]

Here's some images from the test with less photos but higher settings
http://img841.imageshack.us/img841/5306/d2photoscan.jpg

http://img718.imageshack.us/img718/5122/d2zbrush.jpg

slightly more detail made it into the mesh,but not a great deal considering how long it took to build...i also masked all the photos before build stage this time which seems to have helped a bit...but took me ages to do.

Cheers,
mala

[gjpetch]

]A test scan of a gracious workmate, using a nikon dslr this time, test renders in vray with fastsss2.
The raw scan had a few more artifacts that I quickly smoothed out by hand.

6014

6015

6016

6017

[mala]

Holy sh*t!:notWorthy:

well you've put my efforts to shame..that is amazing gary.
Do you mind giving some info?
photo size,number of photos and build settings in photoscan?

what method did you use for the texture build?

I can only hope i can get close to that level of sharp detail.
Cheers,
mala

[gjpetch]

Cheers mala, I used 27 photos with the default settings in photoscan. I think the quality of your scans is probably as good or better than my example! The only difference is that I spent a few minutes smoothing out bumps in sculprtis, and I used a photoshop high pass filter on the texture map from photoscan to generate a bump map. It sounds like such a hack, but it works pretty well for skin (so long as the lighting isn't too directional). Generally all the fine details in skin that show up after a high pass filter will be pores and wrinkles, and they naturally self-shadow, so they will show up darker, making the end result usable as bump or displacement.

Such cool stuff, I've wanted a tool like this for years. now I'm hanging out to go and take lots of photos of gnarly tree roots and sandstone walls and stuff.

[mala]

After i posted i did wonder if you had done that, like you did with the hand you posted,Well it still looks great!

I haven't done any work on my scans in ZB yet just because i'm trying to tweak the photos and settings in Photoscan to try and get as much out of it at each stage.

I set another one going last night with 20 photos and slightly higher settings,which has pushed the detail up a bit.
I'm thinking that using as many pics as possible to align, which is a fairly fast part of the process is a good idea.Then use less photos and masks for the build,then maybe even less photos for the texture build(with masks again)...this way i seem to end up with a sharper texture when using the Average mode.

mala

ps.sorry for your getting your name wrong Greg,it was a long day:)

[oglu]

is this working with the standard version of agisoft or do i need the pro..`?

[Cutia]

HEy guys, happy to see you all playing with photoscan, I was amazed when i found about it too.

This week I will try to do a photo session on a studio with pro lights, i think it will help.
With a good light rig i can stop down the lens and minimize the DOF, i thik that is why i always get crap noses.

Did anyone could complete a full head? whit the jaw and top of the head?
those two parts always give me trouble no matter how much pictures i take.


@oglu

The standard.
It is the one that i have, and works very fine.

[andreicirdu]

:) Interesting little piece of software.

tools used:
agisoft
29 photos taken with mah nokia cellphone (3 MP)
mah face
a hat with 2 "marker cards" ( it helps because the photos are crappy)

http://www.cgfeedback.com/cgfeedback/attachment.php?attachmentid=6068&d=1301847147

http://www.cgfeedback.com/cgfeedback/attachment.php?attachmentid=6069&d=1301847153

[Cory]

Awesome stuff Merry!

[moonjam]

Having a LOT of fun with PhotoScan - incredibly impressive software. Still working out the best approach when lighting/shooting the subject matter but what it can create from just photographs is jaw dropping.

Have done a few tests so far with different setups - this was shot from a fixed position with the skull sat on a turntable inside a light tent.

http://moonjam.com/CGFB/Skull_Scan_005.jpg
http://moonjam.com/CGFB/Skull_Scan_003.jpg
http://moonjam.com/CGFB/Skull_Scan_002.jpg

Spent an hour doing a quick & dirty retopology, would still need more work to be usable but fun none the less! :D

http://moonjam.com/CGFB/Skull_Scan_001.jpg
http://moonjam.com/CGFB/Skull_Scan_004.jpg

[ulf]

cool results! and nice renderings!

[meshmasters]

Great job, very impressive!

Regards,

Joe

[Christian Fischer]

cool results :)

did maybe anyone try scanning a head with agisoft photoscan using multiple cameras?

[Infinite]

...came across this site today http://digicave.com/home/ some interesting work. All shrouded in mystery of course, which is always the way with most scanning services and R&D specialists. I have a feeling they are using http://www.4dviews.com/

http://www.ir-ltd.net/uploads/Digi_1.jpg

http://www.ir-ltd.net/uploads/Digi_2.jpg

Interactive model

http://digicave.com/portfolio/fashionandretail

[moska]

hi
really interesting stuff

[Cory]

Here's an artist that worked on the matrix sequels, he shows some of the work in the link. It's really impressive what skin shaders/models they achieved back in the day.

http://www.psychoform.com/htmlgallery/matrix.htm

[Infinite]

Thanks Cory for sharing these, I have seen all these before many times but not in such high res. Amazing what the big studios were able to do with all that money and such limited experience back then.

[gjpetch]

After some more experimenting with agisoft photoscan, the best results I've gotten are from very diffuse objects, with mottled colouration. Rocks and trees work really well.
I tried a lifecast with mottled paint roughly brushed all over it, and I think the results are pretty great, especially considering I used a point-and-shoot camera, and a bedside lamp for lighting. 90% of the weird lumps aren't artifacts, but rather problems in the original lifecast.

[Vimmy]

Damn! very nice stuff:) Lovely hand to

[Lapaev]

gjpetch, crazy results :)

[Infinite]

Very cool! great hand capture ! :beerchug:

[misha]

Here are some photo scan results from just a handful of picture sets I took about 10 years ago, with just 3 to 5 pictures for each model.... as a note, The Pieta, by Michelangelo, was about 30 feet away, behind non-glare glass...

This software is wonderful, I'm excited to do more higher-quality tests...

Let me know what you think...

Cheers,
Mike

[NextDesign]

Hey everyone. I've been working on 3d scanning techniques for about 3 years now. I've successfully created 3 different methods of capture, all passive techniques.

One of the ways is done by using Structure from Motion to determine the 3d transformations of a camera. Once all the camera positions have been determined, features are detected in each image, and then triangulated with those camera positions.

The second uses stereo photogammetry to create a disparity map. Both of these techniques require no calibration. The nice thing about the techniques I've figured out is that they only call for one camera.

Here's an example of the first technique (This is 1/4 the supplied resolution. I actually don't have enough RAM to process at full resolution :p):
http://i24.photobucket.com/albums/c30/NextDesign/ScanResult.png

And here's the setup:
http://i24.photobucket.com/albums/c30/NextDesign/DSC01514-1.jpg

http://i24.photobucket.com/albums/c30/NextDesign/DSC01517-1.jpg

The first technique I created called for a projector and two cameras. What happens, is a point is projected onto an object. The two cameras capture this projected point, and then triangulate that projected point. It was VERY slow though.

I'll be keeping an eye on this thread. I'm always interested in new techniques! One thing you guys might be interested is this forum for photogammetry. They have a large list of available services and tips for making your own. http://www.pgrammetry.com/forum/

[Infinite]

Hey everyone. I've been working on 3d scanning techniques for about 3 years now. I've successfully created 3 different methods of capture, all passive techniques.

One of the ways is done by using Structure from Motion to determine the 3d transformations of a camera. Once all the camera positions have been determined, features are detected in each image, and then triangulated with those camera positions.

The second uses stereo photogammetry to create a disparity map. Both of these techniques require no calibration. The nice thing about the techniques I've figured out is that they only call for one camera.

Here's an example of the first technique (This is 1/4 the supplied resolution. I actually don't have enough RAM to process at full resolution :p):

The first technique I created called for a projector and two cameras. What happens, is a point is projected onto an object. The two cameras capture this projected point, and then triangulate that projected point. It was VERY slow though.

I'll be keeping an eye on this thread. I'm always interested in new techniques! One thing you guys might be interested is this forum for photogammetry. They have a large list of available services and tips for making your own. http://www.pgrammetry.com/forum/

Hey John, thanks for posting.

We moved this post to a more relevant thread with regards to general 3D Scanning and research. Your post and results are very interesting. It would be cool to see more detail processed captures. That's the hardest part I find. Photoscan seems to produce similar results as yours, so interested to see how far you can push it.

Also that forums is great. It would be cool to see a Scanner-Killer dedicated thread over there as well as a DI3D, DI4D one.

[Infinite]

Here are some photo scan results from just a handful of picture sets I took about 10 years ago, with just 3 to 5 pictures for each model.... as a note, The Pieta, by Michelangelo, was about 30 feet away, behind non-glare glass...

This software is wonderful, I'm excited to do more higher-quality tests...

Let me know what you think...

Cheers,
Mike

Cool stuff Misha :)

[Grandizer]

gjpetch - hand looks awesome .. with normal cam .... need to try this :D

Lee -this thread is awesome man .....thx for all the info

[Hobsons]

Can anyone recommend a budget workflow for stitching agisoft scans together and exporting the colour? I get out of memory errors when trying to do too multiple chunks.

I've tried meshlab, but the model gets softened a lot and I lose the texture info.

Is DI3Dview from the beginning of the thread expensive?

[misha]

the problem with photoscan is you can now reproduce exactly any sculptural work of art from photos....

Cheers,
Mike

[Infinite]

Thanks Furio!

Can anyone recommend a budget workflow for stitching agisoft scans together and exporting the colour? I get out of memory errors when trying to do too multiple chunks.

I've tried meshlab, but the model gets softened a lot and I lose the texture info.

Is DI3Dview from the beginning of the thread expensive?

DI3Dview is free but it has no tools (plug-ins) Most tools cost around £5k to £10k, Shapetransfer, Morph etc you can purchase them in 3 month instalments but it's very costly.

I think there are other softwares out there that can help, not sure what to recommend.

[jonathan]

Hi, My first post here. Ive been watching this thread for a while now and and am trying to get to grips with PhotoScan with varying success. Im struggling to find a workflow that allows the texture atlas to be used on a retopologised 3ds max model. Has anyone had any experience with this? I'd be really interested. Here is a link to my progress to date as part of a larger project: http://www.tidalpoints.wordpress.com

[gjpetch]

Jonathan, I've gotten good results using the projection modifier in 3dsmax to transfer the texture atlas to a retopologised mesh, and also to generate displacement maps. Unfortunately the projection workflow in max is pretty unintuitive.
Basically, give the high mesh a self illuminated material with the photoscan texture in it's diffuse, apply a projection modifier to the unwrapped retopologised mesh (above a turbosmooth), in the projection modifier pick the high mesh, and push the cage out until it covers the high mesh. Then go render>render to texture, enable projection, under options pick a min and max height, add a complete map and a height map, pick a resolution and path, hit render, and done! It sucks, but it does work, best of luck with it!

[jonathan]

Hi Greg, thanks for your reply. I shall certainly give it a go in the coming days. How is the quality of your displacement map from the texture atlas? I had thought of using the high poly model itself as a render to texture displacement map? Thanks again

[MDB101]

Did u guys see this yet?

http://labs.autodesk.com/utilities/photo_scene_editor/

[Infinite]

Did u guys see this yet?

http://labs.autodesk.com/utilities/photo_scene_editor/

Hey I just moved your post here as it is more relevant to this thread. It certainly looks like some very cool tech indeed! Is it out now?

[MDB101]

Hmm dunno, just found it today. I wanna try it though hhehe

[MDB101]

http://labs.autodesk.com/utilities/photo_scene_editor/overview/


The trial version is out now, ill give it a shot tonght :P

[Infinite]

http://labs.autodesk.com/utilities/photo_scene_editor/overview/


The trial version is out now, ill give it a shot tonght :P

Cool interested to hear how you get on:thumbsup:

[MDB101]

Bah was not able to connect to there servers at Photofly 2.0, says connection error, unable to connect. I tried so many ways lol, ill write them for this problem.

Another thing i found was faberic micro scanning? seems like over kill but lol looks good to me ;)

VID
http://vimeo.com/23132440

[Infinite]

Bah was not able to connect to there servers at Photofly 2.0, says connection error, unable to connect. I tried so many ways lol, ill write them for this problem.

Another thing i found was faberic micro scanning? seems like over kill but lol looks good to me ;)

VID
http://vimeo.com/23132440

That's unlike Autodesk :banghead: kkep trying :thumbsup:

That fabric research link is amazing, the surfaces look ace!

[morbo]

btw...have you guys checked out project photofly 2 by Autodesk?

http://labs.autodesk.com/utilities/photo_scene_editor/

its pretty badass for a freebie :)

[Infinite]

btw...have you guys checked out project photofly 2 by Autodesk?

http://labs.autodesk.com/utilities/photo_scene_editor/

its pretty badass for a freebie :)

Hi Harry, I moved your post to this thread as its more related.

Yes it looks very cool, although some of the results if not taken well can be a little distorted, also very low resolution. But an elegant solution. Only other drawback is it's all stored and arranged on a server, so they must get access to allot of free material themselves, I would imagine it also restricts what you can and can't process, subject matter wise.

Can't wait to see how this progresses in the future :thumbsup:

[GregS]

I been testing photo scene editor on set:

http://www.gregstrasz.com/wordpress/wp-content/gallery/selfless-a-tale-of-shade/studio_v0013.jpg

http://www.gregstrasz.com/wordpress/wp-content/gallery/selfless-a-tale-of-shade/vr_gs_v002.jpg?961696894&width=640&height=564

Is really impressive what you can do in 20 min...

[moonjam]

That's excellent Greg!

I have a photoshoot coming up & I'm going to try to catch as much data from the set as possible, good to see it works! :D

[Infinite]

I been testing photo scene editor on set:

Is really impressive what you can do in 20 min...

Very cool! it would be ace to see the wire frame of that room Greg : ) or maybe even a textured fly through!

[pitiwazou]

My test with photofly !

http://uppix.3dvf.com/images/cedricxkx.jpg

Very simple soft ! I really like it.
For a best quality, we must take much more pictures, and zoom for details areas.
Cloud computing is good too !

[Infinite]

My test with photofly !

http://uppix.3dvf.com/images/cedricxkx.jpg

Very simple soft ! I really like it.
For a best quality, we must take much more pictures, and zoom for details areas.
Cloud computing is good too !

Cool Cédric, hope to see some more tests like this! :)

[pitiwazou]

Next week, I will make a 3D version of a french actrice, so I will show you the scan and the final, render on Guerilla of course ^^

[Intervain]

that's some pretty good results with that soft clap

[MDB101]

This guy here knows what hes doing regarding photofly / zbrush. Check out his videos he posted on this thread.



Name: Joseph Drust

http://www.zbrushcentral.com/showthread.php?t=139382

[pitiwazou]

Thx fo the link, it's very good ! I will try this weekend !!

[michal81]

Hi all. Great thread. I'm researcher in the field of computer vison. I'm interested in your experiments. Our research team is developing a Structure From Motion pipeline:

http://ptak.felk.cvut.cz/sfmservice/websfm.pl?menu=webservice

In particular, I'm focusing on dense 3D reconstruction from multiple images:

http://ptak.felk.cvut.cz/sfmservice/websfm.pl?menu=cmpmvs

I'm starting to play with using kinect as an texturing device as well. I'll post some results later...

[Infinite]

Hi all. Great thread. I'm researcher in the field of computer vison. I'm interested in your experiments. Our research team is developing a Structure From Motion pipeline:

http://ptak.felk.cvut.cz/sfmservice/websfm.pl?menu=webservice

In particular, I'm focusing on dense 3D reconstruction from multiple images:

http://ptak.felk.cvut.cz/sfmservice/websfm.pl?menu=cmpmvs

I'm starting to play with using kinect as an texturing device as well. I'll post some results later...

Hi Michal, I moved your post here as it is more related to this thread. Please post some more on your findings, your research looks very interesting.

Lee

[collings]

here is a test i did this week-end with photoscan. Around 12-20 pics with smooth and high settings.
8507

[Vimmy]

Interesting result nic:)

[collings]

thanks sam :) definitely works better without the flash haha, ;)

I took a new set of pictures on the rooftop with overcast lighting. Result should be a bit better i think.

[misha]

scanning using the kinect

http://3.bp.blogspot.com/-t45jFKpKqag/TmFhWxuwFzI/AAAAAAAAAEg/XNT4fHosI_0/s1600/bscan3.jpg


http://1.bp.blogspot.com/-G2DzTlslXrU/TmFhWuEbFOI/AAAAAAAAAEY/sJK7Tx3OJ0E/s1600/bscan2.jpg


http://3.bp.blogspot.com/-xRoFDl1T1ZM/TmFhXK-PB3I/AAAAAAAAAEo/7d_jcndYYGc/s1600/bscan.jpg


this device is great....scanning for less then $100...

Mike

[Rox]

Hi here's my test i did using agisoft photoScan. I scanned Minas Tirith miniature.

I shot 18 photos. The settings are set to fast-medium. Unfortunately my computer(only 2GB of RAM..) can't handle higher settings, that's why the details are lost. I wish i had done more photos from the back of the model... It would look better. I will redo this test once I get a new pc. :)

cheers


8578
8579

[companioncube]

Hi here's my test i did using agisoft photoScan. I scanned Minas Tirith miniature.

awesome! how big is the minature?

[Rox]

awesome! how big is the minature?

Thanks.
It's about 5" high and 4" wide. Of course the miniature isn't made by me. It's from the Collector's edition of Return of the King. ;)

[WhyteKnyte]

hey infinite, will you be releasing any of these models? Like the dawson one?

[Infinite]

hey infinite, will you be releasing any of these models? Like the dawson one?

Hey Mark, no I wont be releasing those but have others lined up.

[WhyteKnyte]

Hey Mark, no I wont be releasing those but have others lined up.

ha ok thanks anyway Lee. Cant wait to see what you have lined up :)

[Ten24]

Hey guys. I don't post on here very much I should probably start taking part a bit more ;)

Anyway I hope you don't mind me posting a few of my attempted full body Artec Scans. There is no clean up on these at all apart from the last image which has been retopologised very quickly in Zbrush.

They were scanned using a mannequin and a turntable with an Artec M (black box version)

I'm going to try it on some real people tomorrow ;)

I have to say a massive thankyou to Lee for all his support and advice over the past year or so.

Cheers

Jamie

http://www.ten24.info/_jamie/stuff/Fullbodyrender.jpg

http://www.ten24.info/_jamie/stuff/Fullbody.jpg

http://www.ten24.info/_jamie/stuff/waterproofs.jpg

http://www.ten24.info/_jamie/stuff/Trousers2.jpg

[Infinite]

REALLY cool results Jamie! The rendered version looks great. Can't wait to see how you get on with full bodies |sculptor| I hope your M tests turn out better than my L tests. I never got around to trying out some M Scans. Maybe think about a quick sweep pass, to get the full volume first then a slow detail pass. I saw guys overseas who got some ace results using that technique - http://www.3lateral.com/ they never posted anything about it though. It looked amazing. They used an MHT like Artecs examples.

Also perhaps knock out your subjects and hang them from the ceiling? That way they might stay still :rofl:

Would be great to see you post some more here! :bouncy:

[meshmasters]

Hi,

Wow, that looks outstanding, great job!

Cheers,

Joe

[Ten24]

Thanks Lee when you say do a quick pass then a high detail pass do you mean 2 separate scans? or just go over the subject once then again much slower? I've not tried that before. The problem I always run into is under and around the arms. I need to do the scans in an arms out pose it always seems to lose registration when I try scanning the underside of the arm.

Im way ahead of you, I've been working on a new type of nerve toxin that will render a person totally motionless but completely rigid. I've been testing it on the guys in the office but they are getting a bit tired of waking up at 3 in the morning in a tpose :)

Yep Ill definitely post here more often by far and away the best CG forums on the net

Cheers Joe :) I'm glad you like them, you should post up some of your Artec scans I saw your automated system on FB a while back looks amazing!!

[PascalR]

Great work James, thanks for sharing !

[Infinite]

Thanks Lee when you say do a quick pass then a high detail pass do you mean 2 separate scans? or just go over the subject once then again much slower? I've not tried that before. The problem I always run into is under and around the arms. I need to do the scans in an arms out pose it always seems to lose registration when I try scanning the underside of the arm.

I haven't tried it myself but like a real quick pass to get a rough shape of the volume of your subject (no worries about a few holes), quick sweep up and down and around. Then the next scan, really slow, get as much info as possible. You then use the first pass as the base registration model to align the detail pass too. In theory it should help.

Arms are a big issue, due to movement and wobble. It's why you see most scans on-line with their hands in pockets or arms crossed.

Maybe a turntable with built in arm holders, poles that come up to hand height when standing? Arms out is never practical unless you have a Cyberware behemoth.

Im way ahead of you, I've been working on a new type of nerve toxin that will render a person totally motionless but completely rigid. I've been testing it on the guys in the office but they are getting a bit tired of waking up at 3 in the morning in a tpose :)

:rofl:

[BrettSinclair]

Those waterproofs look insane. haha

[Ten24]

Thanks Pascal ;)

Cool I will give it a go tomorrow sounds like it should work, We were actually talking about building a motorised turntable with two adjustable arm holders today. It's definitely worth a try

Yea they are a bit crinkle over kill Brett ;)

[NextDesign]

Have you guys seen this yet? Very cool!

http://www.youtube.com/watch?v=quGhaggn3cQ

[Infinite]

Have you guys seen this yet? Very cool!

http://www.youtube.com/watch?v=quGhaggn3cQ

Hi John, I moved your post as it is more relevant to this thread. Yes that is very cool. The next version of the Kinect will be amazing.

[Ten24]

The connect stuff looks very cool, I cant wait to see what happens when they stick a higher res camera in there. I wonder if Microsoft will release some scanning software for it?

Here is a very (I mean very) quick test matching the mannequin clothing scan to a pre existing model. I guess the next step is to retopologise the scan data to create something that is actually useful. That big crease on the back will have to go.

http://www.ten24.info/_jamie/stuff/Body.jpg

[Infinite]

The connect stuff looks very cool, I cant wait to see what happens when they stick a higher res camera in there. I wonder if Microsoft will release some scanning software for it?

Here is a very (I mean very) quick test matching the mannequin clothing scan to a pre existing model. I guess the next step is to retopologise the scan data to create something that is actually useful. That big crease on the back will have to go.

http://www.ten24.info/_jamie/stuff/Body.jpg

Very cool, works well.

Did you get a chance to try some full body captures?

[Ten24]

Cheers Lee

Yea I tried a few this morning but they didn't come out too well I was experimenting a bit with the scan setting and it all screwed up a bit and by that time I'd ran out of "Can I please scan you and can someone else please spin this turn table for me" credits. I think I need to find another room to do these tests in I'm not sure how much the guys in my office like having flashing bulbs in their face all day :)

[Infinite]

Cheers Lee

Yea I tried a few this morning but they didn't come out too well I was experimenting a bit with the scan setting and it all screwed up a bit and by that time I'd ran out of "Can I please scan you and can someone else please spin this turn table for me" credits. I think I need to find another room to do these tests in I'm not sure how much the guys in my office like having flashing bulbs in their face all day :)

Good looking nude models for the win!! Male or Female depending on your preference of course.

[Ten24]

Hey guys here are some more scan tests this time I've cleaned up the data a bit and retopologised and matched it to a body I already had. The head is a scan too. I should texture it all properly but I don't really have the time at the moment

http://www.ten24.info/_jamie/stuff/Colour_Big.jpg

http://www.ten24.info/_jamie/stuff/Colour_Spin.jpg

http://www.ten24.info/_jamie/stuff/Grey_Spin.jpg

[Infinite]

Very cool Jamie, the scan look great. Nice render, was it Lightwave 10? they fit well on to Raf's sculpt.

[BrettSinclair]

Really cool stuff James. Man I'm feeling obsolete these days, Thanks haha.

[Ten24]

Thanks Lee, no it was just LW 9.6 I haven't really switched properly to 10 yet, yea it didn't take much to fit those clothes to Rafs model.

Ha ha don't worry Brett I'm pretty sure these scans are only every going to be useful as reference I doubt its ever going to take the job of a modeller.

[Ten24]

Hey again

Here is a quick spinny of the model for anyone who is interested ;)

http://www.youtube.com/watch?v=okueX4xAZ18

Cheers

[Rox]

Looks very cool James. Mind if I ask, what config do you have to handle so many poly? And how much time does scanning process take? Thanks

[Ten24]

Thanks Rox

I just have a standard (ish) computer nothing special. The scanning process takes about 30 seconds per scan then there is the processing using the Artec software which aligns and merges the scans and a certain amount of clean-up. The rest of the clean-up and retopology was done in Zbrush all in all I think this entire project probably took about 4 days for an untextured model.

[Infinite]

Hey Jamie, it looks very cool although it's a bit misleading to call it a full body scan which it isn't really. The clothing is scanned from a mannequin no? then a separate head, attached to a hand sculpted body done by Raf?

It looks very nice in the spiny though :thumbsup:

**Edit re-read the title on the video, I see you put the info about the arms and such there. Might be nice to credit Raf : )

[Ten24]

Yea you are right Lee the body isn't a scan well the arms and feet, but like you say I did mention it in on the intro screen ;) and yea your right the shirt is scanned from a mannequin and combined with Chris's trousers (not a mannequin, trousers don't sit right on them). I don't see any reason why you couldn't do a 3 part full body scan though using the same process its no where near as quick or as accurate as your scanner killer system though. I just need someone with enough patience to stand on a turn table and let me try scanning them time after time ;) Ill stick a credit in for Raf.

[Infinite]

Yea you are right Lee the body isn't a scan well the arms and feet, but like you say I did mention it in on the intro screen ;) and yea your right the shirt is scanned from a mannequin and combined with Chris's trousers (not a mannequin, trousers don't sit right on them). I don't see any reason why you couldn't do a 3 part full body scan though using the same process its no where near as quick or as accurate as your scanner killer system though. I just need someone with enough patience to stand on a turn table and let me try scanning them time after time ;) Ill stick a credit in for Raf.

Perhaps full "Mannequin Scan" is more accurate :) I reckon you could do a good full body with an Artec.

I believe Artec-Group built their own motorized turntable for the examples on their site, could be very useful. The smaller capture volume with the Artec M should make it possible to do. Interested to see how you get on.

The SK system is good but it still has some noise, it's a very fine balance to get everything set-up correctly. Still on going here.

The guys at Sample&Hold have the right idea, a 4 Mephisto Pod set-up but your looking about £80k-£100k right there.

[Ten24]

Its only the T-shirt that is a mannequin the trousers and head are real person. its a sort of half and half.

Yea it can definitely be done just need a bit of time to perfect the process. I think your SK stuff looks as good as the Sample and hold Stuff I'm sure they probably smooth a bit of noise out of their models before they show them off?

I didn't realise Artec did a turn table I might have a look into that.

[Infinite]

Its only the T-shirt that is a mannequin the trousers and head are real person. its a sort of half and half.

Yea it can definitely be done just need a bit of time to perfect the process. I think your SK stuff looks as good as the Sample and hold Stuff I'm sure they probably smooth a bit of noise out of their models before they show them off?

I didn't realise Artec did a turn table I might have a look into that.

Ahh very cool, didn't realize the trousers were on a human. The hang did seem much better than a Mannequin, generally the cloths hang so bad on those fibreglass monstrosities..

S&H don't appear to do much sculpt or clean-up at all. If you look at their meshes the alignment is very good and because of the type of capture used by Mehpisto (structured light, fringe patterning) they can acquire higher resolution but I think the colour capture suffers slightly.

[Ten24]

Yea their stuff does look very nice I'm not sure why they arent showing the textures though, I've had a play with a very similar system to the Mephisto and it uses a dedicated DSLR to capture the texture just after the fringe pattern has fired. The data is captured by a pair of machine vision cameras, usually monochrome so they wouldn't really have any affect on the texture. I wonder if they fire all 4 pods at the same time or if they are sequential in which case they might have a problem capturing the texture before the next pattern is off. hmm would love to see their set-up ;)

[Infinite]

Yea their stuff does look very nice I'm not sure why they arent showing the textures though, I've had a play with a very similar system to the Mephisto and it uses a dedicated DSLR to capture the texture just after the fringe pattern has fired. The data is captured by a pair of machine vision cameras, usually monochrome so they wouldn't really have any affect on the texture. I wonder if they fire all 4 pods at the same time or if they are sequential in which case they might have a problem capturing the texture before the next pattern is off. hmm would love to see their set-up ;)

The price comes in the Machine Vision cameras, some new USB3 cameras are coming out in the last quarter which could change the playing field dramatically. Super cheap, super fast and super high-res.

Give S&H a call they are nice guys. To run 4 Pods you need 2 PC's, they are twin synced each I think. I would imagine the DSLR fires either first or on the last frame of the fringe sequence, very fast. I think there are many issues with texture alignment, colour changes which makes the whole colour process a pain. 4DDynamics are another company that need Non-Rigid : )

The system you are referring to is pretty much the same as Mephisto.

[Ten24]

I was speaking to the guys at Artec and they gave me a few pointers on full body captures and this is what I came up with. just 1 scan starting at the top and working down to the feet. worked really well ;)

http://www.ten24.info/_jamie/stuff/FullBodyScan.jpg

[Infinite]

I was speaking to the guys at Artec and they gave me a few pointers on full body captures and this is what I came up with. just 1 scan starting at the top and working down to the feet. worked really well ;)

http://www.ten24.info/_jamie/stuff/FullBodyScan.jpg

Very good Jamie, came out really nice.

[Ten24]

Cheers Lee :) this is straight out of the Artec software (pretty much, I decimated it in Zbrush) so the alignment was pretty good. I think the one thing that really made this work was setting the data compression level to fastest and then just methodically working down the subject on a turn table. It would be even better with 2 artecs ;)

[Infinite]

Cheers Lee :) this is straight out of the Artec software (pretty much, I decimated it in Zbrush) so the alignment was pretty good. I think the one thing that really made this work was setting the data compression level to fastest and then just methodically working down the subject on a turn table. It would be even better with 2 artecs ;)

What do you mean by Data Compression level? I agree more Artecs is always a good thing :)

[Ten24]

What do you mean by Data Compression level? I agree more Artecs is always a good thing :)

This little setting here ;)

http://www.ten24.info/_jamie/stuff/datacompression.jpg

[Infinite]

This little setting here ;)

http://www.ten24.info/_jamie/stuff/datacompression.jpg

Ahh that's new. I barely use the Artec these days and hadn't noticed that in the new update. I guess it some how reduces the quality of each captured frame?

Although I don't quite get that? Maximal Speed? or Maximal Compression? surely faster capture speed would mean more Compression?

[Ten24]

Yea it doesn't make much sense but if you stick it up to maximum speed the scanner is much quicker and seems much less inclined to lose registration. I didn't notice much of a drop in quality on the captured frames I did process this with the fusion resolution set to 1.5 though so its a bit smoother than the usual 0.5 that I use but it seems to work just fine ;)

[Infinite]

Yea it doesn't make much sense but if you stick it up to maximum speed the scanner is much quicker and seems much less inclined to lose registration. I didn't notice much of a drop in quality on the captured frames I did process this with the fusion resolution set to 1.5 though so its a bit smoother than the usual 0.5 that I use but it seems to work just fine ;)

Very cool thanks for the extra info Jamie.

[Infinite]

I have been seduced by the dark side! Love Agisoft.

Really impressive software.

0.01 Capture time, allot of Cameras. Full Body (1 angle) with built in high-res face capture

http://www.ir-ltd.net/uploads/cgfeedback/Lee-Agi-01.jpg

http://www.ir-ltd.net/uploads/cgfeedback/Lee-Agi-01-close-up.jpg

[NextDesign]

I have been seduced by the dark side! Love Agisoft.

Really impressive software.
Yes, their software is great. I've been playing around with their StereoScan program with my 5D.

Here's a result, which has been smoothed a bit, as I shot it outside and had some shadow artifacting:
http://i24.photobucket.com/albums/c30/NextDesign/photoscanResult.png

[Infinite]

Here are some more Agisoft tests. A completed 360 Full body capture.

Captured using a motorized turntable and lots of synchronized cameras.

http://www.ir-ltd.net/uploads/cgfeedback/Lee-360-01.jpg

[NextDesign]

Very nice mate. How many cameras in total where there? Were you using active projection as well?

[Infinite]

Very nice mate. How many cameras in total where there? Were you using active projection as well?

Thanks. Over a dozen DSLR's on a fixed rig. Pure Passive no projection or structured light.

[moonjam]

That's absolutely incredible Lee. Would love to see some shots of your setup - although I appreciate that you might not have a spare camera to take a photo!

[Infinite]

That's absolutely incredible Lee. Would love to see some shots of your setup - although I appreciate that you might not have a spare camera to take a photo!

Thanks AJ. No I'm out of Cameras! I will try and post a picture later in the week though.