Messages

1

General / Re: Ikea Lack Photogrammetry Turntable

« on: July 11, 2018, 08:32:49 PM »

Thanks for sharing!

No problem. I had documented process anyways so I could make another if need be, or if I screwed up. 

Where did you get the lazy susan from?

Links to the items purchased are in the article. The Lazy Susan is from Amazon. It was a TamBee brand.

2

General / Ikea Lack Photogrammetry Turntable

« on: June 11, 2018, 07:31:40 PM »

Thought somebody might be interested in this DIY info. I just put together an article post going over the steps I took to convert an old Ikea Lack table into a small object capture station. Complete with Azimuth circle, 22" lazy susan, and a glass top to allow for bottom capture of items.

There are tons of photos showing the process and links to materials used.

You can view the 3 page article here:
https://www.provideocoalition.com/ikea-lack-photogrammetry-turntable/

3

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 10:53:17 PM »

Apologies, I didn't include those initially. Just uploaded a new ZIP to the same dropbox link. The ZIP is called "Gaia_001_TexturingNewRetopo_AllFiles.zip".

4

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 09:58:32 PM »

Alexey, Thanks so much for getting this to work. I am curious now, because I didn't realize I had packaged the initial source psz file "gaia-mcap07-pose001-hapd.psz" for you to use. I thought it was just the JPG sources.

I was using a duplicate of that file just so I wouldn't save over it by accident, and called it "Gaia_001_TexturingNewRetopo.psx". The file I sent along in the zip.  "Gaia_001_TexturingNewRetopo.psx" file is the one I was having this issue with. I am now seeing too that there are two different file extenstions psx versus psz. Standard versus pro versions of the software I assume. Could this have caused the glitch I was seeing? A possible corruption of data when duplicating the file?

Regardless, I used the steps as you stated in the email and used the same "gaia-mcap07-pose001-hapd.psz" file instead of my "Gaia_001_TexturingNewRetopo.psx" copy and everything worked just fine as expected. (First image "TextureWorksWithOGProject.JPG")


Your exported OBJ and texture were perfectly fine as well. Quad mesh and the material worked. (Second image inside Cinema4D "StillQuadAndTextureWorks.JPG")

Glad it works now.

5

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 02:15:03 AM »

Just sent off an email with the Dropbox link to you.

6

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 01:00:08 AM »

Same deal. I will say though that the processing time for 8K texture was screaming fast compared to 1.2.6. As was the importing of the mesh.

The 1.3.0_3745 texture result is attached.

I do find it quite eerie that only the eyes are working though, even in 1.2.6. 

7

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 12:41:54 AM »

Good to know, I didn't realize that the display was different than output.

I will get 1.3 and let you know if anything changes.

8

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 12:14:27 AM »

Also here is a PhotoScan import versus Cinema4D import view of the same mesh, so you can see what I'm seeing.

9

Face and Body Scanning / Re: Texture is mangled

« on: February 01, 2017, 12:08:30 AM »

First a small comment regarding the quad mesh, if you are importing the mesh that contain only quad faces and are not removing any polygons in PhotoScan or decimating mesh, then the exported mesh after texturing would still contain same quad faces.

That's what I was expecting, but it didn't. I swore I've imported quad mesh before in the past. It has been awhile though.

As for the texturing, could you please specify if the imported model contain only quad faces or triangles as well?

The imported model should be 100% quads with a clean UV.

I can also suggest to check version 1.3 pre-release to see if the issue is still there, since we were fixing some issues with quad mesh support after 1.2.6 release.

I'd love to try it out. Point me in the right direction and I get it.

I can certainly send you a link to the three OBJ's (Raw, BaseMesh, final output) as reference if that helps too.

10

Face and Body Scanning / Texture is mangled

« on: January 29, 2017, 03:28:56 AM »

This is the first time I've come across this issue. My texture is crazy mangled when using no matter what mapping mode I choose. I really need to use "Keep UV" though.

My pipeline:

- Do standard solve in PhotoScan.
- Export Raw mesh as OBJ.
- Import OBJ into Wrap3
- Import Base mesh from C4D (It is a quad mesh 26k polys)
- Do Wrap3 wrapping function
- Export the new OBJ
- Import the new clean obj mesh into PhotoScan (seems it doesn't like quads and retriangulates the mesh to 52k)
- Run Texture process with Keep UV, Mosaic, and Enable Hole Filling

I attached a screen cap of what I get. Any thoughts on the cause?

11

Face and Body Scanning / Re: Reliance on "features" are too restrictive

« on: June 20, 2015, 03:30:53 AM »

I'm wondering how other people are dealing with this problem, or are they just giving their customers a list of things they cannot wear and also patching up holes in post-production?

Yes, a spec sheet of acceptable materials is given to client, and yes to the other. There will ALWAYS be cleanup after the scan is processed. Hole filling, bump smoothing, texture cleaning, etc....

For some darker fabrics, as shown in this scan I did (https://vimeo.com/104762987), you need high resolution, lots of even clear light, and sharp deep depth of field. You also need to always shoot RAW. Having RAW images can buy back some flexibility when processing the images. For the above scan, and most scans for that part, I run a pre-process of all of the images to brighten the shadows and bump up the exposures without clipping the details, run through Agisoft and once I have a decent mesh, I then swap the images out for more normal exposed versions to then process the texture pass.

12

General / Re: why is my point and shoot making batter captures than a DSLR

« on: June 22, 2014, 02:03:53 AM »

There are a wide number of variables that could cause differences in quality.
1) The location of each captured image is probably not identical between PS and DSLR, therefore resulting in different initial image data for Photoscan to process. This is most likely the main difference.

2) The quantity of images captured. 139 vs 133, also goes back to one set not having the same pixel information as the other.

3) The sensor types and how they capture pixel color information can play a factor.

4) Camera settings, like ISO, aperture, and shutter speed. Can all play a factor in how data is captured.

5) Each camera can reduce and record image data differently to a memory card.

There are many more factors, but what it really boils down to is that you:

- Capture images with a wide depth of field to keep as much of your subject in sharp focus as possible.
- Stereo capture is helpful for the software, so with a single camera you wanna keep each image you shoot within a certain distance of each other. Essentially to where you can take any two photos that are next to each other and easily make a stereo image from it. This require consistent camera location around your subject and will vary depending on the size of your subject.
- Keeping the lowest ISO possible will help reduced grain in the image. The less the grain the better the pixel information for the software.
- Tracking is a HUGE part of it too. If your subject has a smooth, non-detailed surface, then the software has nothing to match in each image no matter how high the resolution. Having some kind of unique texture, be it a printed pattern or in your case a rough material will help in solving the cameras. Ideally having specific markers in the scene work best.
- Coverage, coverage, coverage. This kinda goes back to the stereo pair note above, but making sure as many hidden areas of your subject are captured will help with having a full complete solved surface.
- Avoid macro lenses if possible. They completely destroy any way of getting wide depth of field. It usually chokes it to razor thin space and forces too high of an aperture. High apertures usually above f8-f11 will actually cause diffraction and start softening the image. Sharpness is what you need for the details.

For your two sample images, I can see what looks to be a heavier noise pattern on the DSLR camera. This may also just be the forum compressing the image. What where the camera settings and resolutions used for both capture sessions?

13

General / Anyone use R3DS: Wrap with PhotoScan pipeline yet?

« on: May 27, 2014, 08:31:32 PM »

Has anyone used this yet? I tried the original initial demo many months ago, but was unable to spend much time playing with it. Version 2.1 just released this weekend it looks like and I was curious if anyone has had decent results with a full body human being processed though this. Maybe a useful bridge for the UV conversations. Granted the cleanup process still needs to happen via ZBrush (or insert app of choice), but if this works as simply as it portrays, it looks to be very helpful.
http://www.russian3dscanner.com/


Saw another post, that seems to have fizzled out here when Wrap was released.
http://www.agisoft.ru/forum/index.php?topic=2196.msg11632#msg11632

14

General / Re: Parameterizing texture atlas... hangs at 70%

« on: April 30, 2014, 01:45:22 AM »

Here's a quick screen cap of the final 16k texture of the face. It all seems to be working wonderfully now. 

15

General / Re: Parameterizing texture atlas... hangs at 70%

« on: April 30, 2014, 01:26:00 AM »

Bingo! That seems to have fixed my issue completely. Just did a 4096x1 texture on the 500k poly model and it worked. Doing an 8000x1 right now and it has two minutes left. Much better. Thanks for the help.