Forum

Show Posts

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


Messages - scanlab.ca

Pages: [1] 2 3 4
1
Why is there a mismatch in Maya between virtual cameras and physical cameras?
When I set the Focal Length and Angle of View, the virtual camera sensor size becomes smaller.

Btw, everything works fine, but I'd like to know just out of curiocity. :)


Thank you,

2
Face and Body Scanning / Re: Why is Image Orientation important?
« on: October 19, 2017, 08:23:31 PM »
I see, thanks for the prompt explanation Alexey.

Would it be possible to group calibrations where N and M dimensions could be interchangeable, given the focal lengths are identical?
The reason for the request is that it is just much more intuitive to analyze and mask photos in their preferred orientation.


Hello Ruslan,

If the resolution of original image is N x M pixels, PhotoScan will put all the images to the same calibration group (providing that they are taken with the same focal length). And the image orientation doesn't matter as long as it is just and EXIF tag, but is the camera is applying this orientation information - you'll get two groups: N x M and M x N and the calibration parameters would be estimated separately. Also this sub-division will raise additional confusion for the 90-degree rotated images - whether they are rotated clockwise or counter-clockwise.

So we usually suggest to disable automatic image rotation by the camera. In other words all the images taken by the same camera should width x height resolution and shouldn't be split into two groups: width x height and height x width.

3
Face and Body Scanning / Why is Image Orientation important?
« on: October 19, 2017, 07:07:04 PM »
Hi guys,

Something I'm not understanding correctly probably, but why is Image Orientation important (you know that warning that comes up when adding photos)? Isn't flipped image just a transformed matrix?

One way I see when it matters is only when doing Aerial Photogrammetry and orientation may affect how internal algorithms interpret directional motion blur, but for Face and Body Scanning, does it really matter?

4
Feature Requests / Batch Processing - Start with all items de-selected
« on: January 22, 2015, 06:48:26 AM »
Its a small minor thing which keeps bugging me.

Whenever I want to align chunks Photoscan window starts up with all chunks selected, but for the most part I just want to align a single chunk, not all chunks. and I keep unchecking things.

Would it be possible to make the default behavior where nothing selected?

Maybe some don't agree with me here, so maybe something to discuss?



5
Feature Requests / Import Object in Non-Triangulated State
« on: September 17, 2014, 10:16:19 PM »
I feel like there's a real need for this option.

Would do you say?

6
Feature Requests / Re: Show just lines rather than shading when masking
« on: September 15, 2014, 11:03:04 PM »
This might present another problem where we won't be able to tell which section is masked and which is not.
Maybe something like a Mask Transparency Slider would prove to be useful in such cases.

+1

7
Face and Body Scanning / DigiCamControl
« on: September 15, 2014, 10:57:38 PM »
Hey guys,

After reading a few posts here and noticing that some people are still talking about breeze.
I want to give a bump to the software I've been using successfully to summon images from all of our Nikon cameras to PC. It is absolutely free in compartment to SmartShooter and has a very nice feature set. Also, it seems like the developer will be adding a feature to be able to LiveView all cameras at once! The application was mentioned here on this forum once before a couple of years back, but I'd just throw it out there for anyone new to this.

http://digicamcontrol.com/

Best,
Ruslan

8
Feature Requests / Re: Occluding Contours
« on: September 12, 2014, 03:40:40 AM »
I'm completely satisfied with what we have in PS.
If results don't impress - ignore!

9
General / Re: "Fine" vs. "Normal" quality JPEGs
« on: September 09, 2014, 07:27:22 PM »
If you can shoot RAW then do that and export TIFF's if you can not always use the best JPG setting in the camera, compression artifacts are bad for calibration.

Go with RAW if you use Nikon.

The quality of JPEG Fine vs Raw NEF is at least 1/4, and might actually be even lower.
I was able to tell the difference just by comparing the two when recording both RAW and JPEG Fine simultaneously.

You can always reduce the quality setting in Photoscan to speed up the processing time.
That's what Accuracy setting in "Align Photos" stage is for.
It re-samples your pictures to Half and Quarter at Medium and Low respectively.


10
Feature Requests / Re: PSD Texture Mask Collage
« on: September 09, 2014, 07:09:11 PM »
With Adobe Bridge selected photos can be opened in a single psd flie each on individual layers.

Opening photos is not a problem. The time consuming bit is figuring out which image contributes, in which order and extracting each map individually.

I do have a solution for this, but an out of the box method would've been nice to have.


11
Feature Requests / Occluding Contours
« on: September 09, 2014, 07:02:13 PM »
Not sure if this has been implemented or not, but I'll post this anyway:

http://grail.cs.washington.edu/projects/sq_rome_g2/


12
Face and Body Scanning / Re: Continuous Lighting for full body capture
« on: September 02, 2014, 11:34:53 PM »
There's only 2 cases when you might want to use Continues Lighting for your rig:
1. Keep complexity of your Scanning Rig down, while learning about photogrammetry
2. You have enough cameras to maximize the parallax

Details:
Continues light solutions are way too bright for Character Scanning, since you will need to drop your apertures to f/16 to get enough data. This will become even more of an issue once you go with CPFs.
Having more cameras will enable you to discard more of the peripheral low quality information in each one of your photos: bokeh, lens distortion etc.
Once you maximize parallax and more of your in focus data begins to overlap, only then you will be able to up f-stops to widen up your aperture and let more light in.


I hope this makes sense. :)

13

Thanks for the clarification.   I am afraid that it still does not work for me after trying to follow your tip.
Looking up usemtl it seems to refer to a hint to use a specific material?
I am attaching a file here with my model maybe someone can take a look and see what I do wrong?
Or upload an example .obj that has the udim's working and I could try to copy the syntax?

Cheers
Fred

Oh, I see. I missed a few steps.
Anyway, this G for USEMTL mumbo-jumbo is old.

Just import your model into you package
assign different shaders to each UDIM you have
move UDIMs back into 0-1 UV space
import your model back into Photoscan and you'll have working UV pages
and you'll be able to generate higher resolution textures


14

Hi All,

me too am looking for a way to bring a udim UV tiled model back into PS for texturing.
I am using Maya to make udim tiles and have tried to bring back as .obj.

Scanlab, I tried to look for the tags you mention in the test object .obj that I exported from Maya but couldn't see those so I didn't get it to work.

The file starts with:
Code: [Select]
# This file uses centimeters as units for non-parametric coordinates.

g default
v 4.232166 89.589348 38.481514
v 4.319020 39.931038 46.131218

and after the 'vn' (normals?) entries I can see,
Code: [Select]
vn -0.021015 0.152176 0.988130
s off
g polySurface1
f 5/1/1 6/2/2 1/3/3 2/4/4
s 1


Could anyone that managed to get this to work explain some more?


Cheers
Fred (Photoscan newbie)

"g polySurface1" is the group i'm talking about,
if you were to do search&replace on "g polySurface*" and change that to "usemtl polySurface".

Then you'd get you UV pages working properly.


Best,
Ruslan

15
Feature Requests / Re: Texture to Vertex Color
« on: May 26, 2014, 06:09:06 PM »
The PLY export does not work. I'd like to be able to convert high resolution texture to vertex color.. a hackish pTex I guess.

 

Pages: [1] 2 3 4