Topics

1

Python and Java API / region box translation after re-oriented

« on: July 15, 2015, 06:17:20 PM »

i'm confused about the local and re-oriented coordinates math

my body scanning workflow detect markers on ground and then update the transform. This makes a [R] on the chunk name and result in correct alignment, orientation and scale of scene.

however, now i want to move the center of region box to [0,100,0] (by script)   (100cm above the ground). when i do, it still works on a local coordinates, instead of the new re-oriented coordinates.

any help on the correct code is appreciated.

2

Python and Java API / region center

« on: May 12, 2015, 11:01:09 PM »

hi

I have a referenced chunk
i want to translate the region center to a marker position
       
        newregion = chunk.region
        newCenter = chunk.transform.matrix.mulp( chunk.marker[0].position )   
   newregion.center = newCenter
   chunk.region = newregion
       

the result is wrong. is it my code is wrong?
thanks

3

Face and Body Scanning / classify ground for human scan

« on: May 06, 2015, 04:36:32 PM »

is it possible to identify the floor by using classify ground automatically?
the scene i do is aligned and scaled, however, i tried with varying parameters, photoscan just freeze and calculate 0-10% forever.

thanks

4

Python and Java API / bounding box region

« on: May 06, 2015, 12:56:44 PM »

hi,

is there a way to code the following simple idea?

after aligning cameras and aligned the scene by coded markers, i want to specify a hardcoded coordinates of region,
e.g. [1,2,1] for a fixed volume of 1m wide x 2m height x 1m depth for scanning human

i tried some
newregion.center = PhotoScan.Vector([0,0,0])
newregion.size = PhotoScan.Vector([1,2,1])
chunk.region = newregion

but the region is Off
and secondly i don't know how to reset the rotation to zero ( align to the 3 axis)

advise is appreciated

5

Python and Java API / import marker coordinates

« on: May 01, 2015, 06:26:16 PM »

hi

I'm trying to python the process of detecting markers of coded targets, and then import the predefined coordinates of markers exported from previous sessions of GUI (same markers)
the following code runs and detect markers works but then it doesn't seem to assign coordinates values to the markers. (where i successfully did so from GUI)

        chunk.detectMarkers()    
   chunk.loadReference( markerFile, "csv")
   chunk.updateTransform()   

where #markerFile is the full path of the CSV .txt


secondly, chunk.detectMarkers( type = TargetCircular12bit, tolerance=50 ) gives me errors so i stick with default without parameters but i hope i can have it


any insights appreciated!

6

Python and Java API / set focus to a chunk

« on: April 29, 2015, 10:23:46 PM »

hi,

in python, i would if i can set focus to a newly added chunk (make it the active chunk before processing, so that i can see the process update as it goes and visually detects any problems in the middle)

thanks a lot!

7

Python and Java API / replace photo

« on: April 29, 2015, 10:14:47 PM »

hi

im scanning human
i have been trying and searching but i can't seem to find the right python to do the "replace path" GUI function that i needed to replace photos with projection to photos with clean texture

code:

for i in range (len(chunk.cameras)):
      camera = chunk.cameras
      camera.photo.open( camera.photo.path.replace( chunkNum_proj, chunkNum_calTex) )

chunkNum_proj and chunkNum_calTex is before/after folder names

the above code runs but then the texture is black and when i double click on any of the photos to check if it's replaced, it pops up a dialog "can't load image:" { path of new image}   (but the path is correct)


any advice/ sample code appreciated!

8

General / import cameras basic understanding

« on: April 13, 2015, 07:06:40 PM »

hi,

any insight on how to use "import cameras"?

Im scanning a quantity of human with fixed array of cameras.
I wish to accelerate the processing time
so i did:
1) mask by background
2) use OpenCL card for dense point cloud

I'm wondering about the import camera functions which i never use, and is greyed out after i imported the photos. i just don't understand how it works.

I was thinking: if i align photos for the first scan, then export cameras, since the cameras didn't move, then for the subsequent scans, i can just import the cameras without the time consuming "align photos" (and since openCL doesn't accelerate this step), but i can't get it work

any insight on how to use "import cameras". and can it skip the align photos step, is appreciated.
thanks!!

9

Face and Body Scanning / Auto topologize and rig your 3Dscan - Quantum Human

« on: January 26, 2015, 11:34:23 AM »

Introducing Quantum Human
We are excited to introduce our latest Digital Character system:
Quantum Human
AUTOMATIC HUMAN MODELLING & RIGGING FROM ARBITRARY MESH
www.quantum-human.com

We believe body 3D scan users can vastly benefit from it because any A-pose character can be topologized and rigged in few minutes, even you don't have any knowledge on topology and Rigging, your scan will come alive and moving in few minutes!
Besides 3D scans, 3D sculpts e.g. from Zbrush is another great source of input. The system basically accept arbitrary topology and only analyse the shape and then plug everything into your shape.

As you see from the demos in the site, the simplest mode even allows 1-click operation and your character becomes alive.

Sneak Peek Video that inputs a 3D scan model and 1-click operation:
http://vimeo.com/113461164

more videos and features:
http://quantum-human.com/?page_id=496

10

Face and Body Scanning / the new "optimize camera alignment" - is it useful for scanning human?

« on: January 26, 2014, 09:17:36 PM »

My rig is basically only for fully body scan. I found that in Batch processing window, the new (after v1.0) "optimize camera alignment" -
pre-1.0 versions do not have this and i use PS batch to align-dense cloud->mesh without problems.
so will this option automatically improve camera alignment and how should i use it?

thanks for any hints

11

General / pair preselection

« on: November 13, 2013, 08:10:42 PM »

about alignning images
with <120 images, will enabling image preselection (generic) increase or decrease processing time?

secondly, will it make any quality difference to enable or disable it?

thanks

12

Face and Body Scanning / Smooth skin issues and results

« on: November 09, 2013, 08:08:42 PM »

Hi,

see anyone have some insights or similar experience:

i have been scanning a couple of friends' face as test, using my 12 canon rig with a fixed setup of a few 600W flash and f11, iso100. 1/125sec shutter, with cross-polarization setup.

after a few test, im quite convinced that people with smoother skin to less smooth skin/larger pores results in HUGE quality differences:
(not any racial discussion, purely experimental findings) my samples includes Caucasian and Chinese men and girls(without make-up), each group of above contains 2-3 person as samples.
results shows quite clearly that Caucasian men group always produces the highest quality even if i try higher ISO, dimmer flash lights. whereas Caucasian girls (usually with smoother skin than men) producer nosier quality. and Chinese man produces an obviously even noiser result, while chinese young girls produces worst very noisy results. An exception is if i scan a chinese elderly, which producer really nice details with least noise.

reading and trying myself about how Pscan works, i thin kit definitely has to do with different level of skin contrast/ pattern clarity, etc.

i must say visually the causian men scans can withstand full screen head with pleasant looking detail/low noise, while the chinese girls i have to zoom out up to only 1/3 of screen in order to have same "visual quality". such a huge difference.

Then i tried projecting noise pattern (which i need to use lower ISO and wider aperature, slower shutter, sacrifice texture, not ideal), results:
the already high quality Caucasian men scans do not gain significant higher quality results, while others gain increasing level of detail/lower noise, from cauasian girl < chinese men < chinese girls. the latter ones gain huge quality difference and catch up with the caucasian men scans. and this results catch up with the above flash only quality.

but there comes another frustration of acquiring usable texture...

this frustrates me, because i wish to test scan friends of different nationality. While my caucasian men friends are all good without projection noise, others are suffer lower quality, especially chinese groups scans are unacceptable quality without projection.

I tried splitted scan (but i have to manually switch off the flash at the time of projection, and change settings from smartshooter to send to the 12 cameras. this manual work takes me at least 3 seconds , and i suppose this is totally not ideal to ask the subject ot stay still for 3 seconds or above to introduce movement.

i have read this thread repeatedly :
http://www.agisoft.ru/forum/index.php?topic=1324.0
and yet i cant find an ideal way...

any advises?

thanks

13

Face and Body Scanning / cross-polarization

« on: October 27, 2013, 04:47:03 PM »

hi

im experimenting a studio setup of fullbody scan.
i read some web site about cross polarization.
does someone have a detail suggestion on how it can be done properly to give 99% diffuse photos?
1) i try to put the polarizing films on the strobe and camera(with 90 degree rotated direction), however i still got some 40% specular reflection remains. the strobe is directly pointing to the subject. i'd like to check whats done wrong
2) obviously direct pointing the strobe to subject will produce harsh undesirable shadows.  before i come across cross polarization, i used to point the strobe to the white walls and make it a large reflector. can cross polarization works if reflected/diffuse lights like this is used?
 
thanks
viper