Hi, I work in a big game company and we have our internal 3D Scan studio, full body scan and face scan. We use Agisoft Photoscan in our workflow, but we are evaluating the possibility to switch to RC. However, first I’m gonna need to make sure to understand how the software works, and if it has all the features we need, but I’m hitting few walls at the moment.
I’m using the demo, so I don’t know if this is a limitation, or if I’m doing something wrong here.
Our face scan setup uses 38 DSLR, and I drag them all in the main window. Then I click on Align Images, but instead of having all my 38 cameras in one component, I get three different components, one with 28 camera, another with 2 cameras and another with 4 cameras, and the remaining 4 cameras are ignored. Each component generates it’s own piece of model, with it’s own scale and it’s own orientation, which obviously is not what I want.
I want one mesh generated from all my 38 photos, with consistence scale and orientation.
Hi Riccardo Minervino
If you get into alignment issues, first try these settings, set all as here
Alignment_settings_80k.jpg
If it still does not want to align all images properly, then use a sphere with newspaper on it. Set scale, orientation etc and ALIGN it. Then select all cameras, and in ALIGNMENT->METADATA(XMP) use LOCKED ( so that all cameras and parameters are locked )
it will export XMPs next to the aligned images. Then just move them to the other not well aligned images. Load the images in RC ( XMPs are used automatically, so you do not need to set anything ) and ALIGN the bad data…
You need to keep the SAME names for the files and cameras, cannot change the position, orientation … and it will work…
If you get in alignment issues try these settings first, set all as here
Alignment_settings_80k.jpg
If it still not want align all images properly, then use help with sphere with newspaper on it. Set scale, orientation and etc and ALIGN. Then select all cameras, and in ALIGNMENT->METADATA(XMP) use LOCKED ( so all cams and parameters are locked )
it wil export XMPs next to the aligned images. Then just move them to next not good aligned images. Load the images in RC ( XMPs are used automatically , so no need to set anything ) and ALIGN the bad data…
You need keep SAME names for the files and cams cannot change position, orientation … and it will work…
Hi. I tried your settings, and it gets a bit better, i got 33/38 cameras to align. So I used the sphere (we have our own alignment object with markers), and I got all 38 the cameras, but the Metadata (XMP) option is grayed out. I cannot click on it. Is it a limitation of the demo?
Hi Riccardo Minervino
If you want, send me the data and I will process it for you ( milos.lukac@capturingreality.com ) . We have strong reasons why the DEMO cannot export some particular data… You could take 20 demo versions and calculate the whole New York in few days with RC and then just export it in one working version…
There are many customers doing full-body and head scanning with RC, so we know they are happy with the quality and speed compared to other solutions…
If want send me the data and will process it for you ( milos.lukac@capturingreality.com ) . We have strong reasons why DEMO cannot export some particular data… You could take 20 demo versions and calculate whole New York in few days with RC and then just export in one working version…
There are many customers doing Fullbody and head scanning with RC, so we know they are happy with quality and speed compared to other solutions…
Hi Wishgranter, thank you but I cannot send the source material as it’s under NDA. I’m currently negotiating with my Company to purchase a three months license to use as test to continue the evaluation.
Hi Naohiro Yamada
You set 5 units = 5 meters, you need to set 0.05 ( meter ) and then it ALIGN again so that the scale change is included in the alignment.
Then when you export data, again, take care of the PROPER units in MAYA, you need to set MAYA units to meters…
Are you doing that on purpose? I told you I’m Riccardo Minervino, the user name is different because I’m logged in with my company’s account…
Wishgranter wrote:
You set 5 units = 5 meters, you need set 0.05 ( meter ) and then ALIGN again so the scale change is included in the alignment.
Then when you export data keep again care about the PROPER units in MAYA, you need set MAYA units to Meter…
That doesn’t change anything. I did exactly what you wrote, and the model is still not correct. It seems to be about 4-5% smaller than it should be.
Also, the orientation is wrong.
I’m gonna give this software one last chance, but it seems it’s too early for it to be used in production.
Hi Riccardo Minervino,
got confused with the names… do we still know who is discussing what ?
That doesn’t change anything. I did exactly what you wrote, and the model is still not correct. It seems to be about 4-5% smaller than it should be.
Also, the orientation is wrong.
Send me the dataset for inspection. 4-5 % should NOT be there, it’s tested quite heavily and verified that the scale is OK. It looks like a workflow issue… Orientation is wrong !!! how wrong ?? any screenshots ?
Hi Riccardo Minervino
I have inspected this image: download/file.php?id=418
and the problem is that you used only ONE image per control point so it could NOT work. You need to place a control point in 2 images, and then ALIGN it to get the PROPER scale…
After alignment look at the CONSTANTS and there is the DISTANCE ( which you have set ) and ACTUAL DISTANCE which defines the actual real measurement…
Try it and you will see it works properly…
Hi Riccardo Minervino
I have inspected this image: download/file.php?id=418
and the problem is that you used only ONE image per control point so it could NOT work. You need to place a control point in 2 images, and then ALIGN it to get the PROPER scale…
After alignment look at the CONSTANTS and there is the DISTANCE ( which you have set ) and ACTUAL DISTANCE which defines the actual real measurement…
Try it and you will see it works properly…
I don’t understand why you say I put the distance points in only one image? I used three different images, or am I missing something? :?
Hi Riccardo Minervino
Look at the first image: download/file.php?id=418
I can see only one image per control point… you need at least 2 images per control point and do not need to create 3 different CONSTANTS
I don’t understand why you say I put the distance points in only one image? I used three different images, or am I missing something? :?
Hello there,
Control Point coresponds to one point in space (on your 3Dmodel). e.g. you got dataset of some statue and you put control point named “point 0” on tip of the nose of the statue in one image, so in the second image you should put same control point (“point 0”) on same place = on tip of the nose of the statue but from different angle.
Problem is you got 3(or more) different control points on same spot on model. If you look on the ruler at 1cm in the image you got there “point 0”, “point 2” and “point 4” that’s 3 different CPs for one point in space.
It should be like this:
CR01_.jpg
But how to add same control point to different images? There are 2 ways I am aware of:
Select CP in one image and then you can add same CP to other image
Drag and drop CP from 1Ds view
And you got another mistake there (I think it is mistake) : distance between 1 and 5 is 4cm not 5cm.
