DIY and Entry-Level 3D-Scanner Forum

first test were very promising. My kittle figurine was replicated very nice. Unfortunately I couldn't save it because of the demo version.
It seems that the quality of the pivtures is important to the reconstruction process. Therefore, I ordered a Canon 40D on ebay
Looking forwar to do more tests with photoscan

_________________
Bernhard
www.virtumake.com

Yes, photo quality is very important to the process. Photoscan can't do its magic from blured, noisy or bad exposed photos. Another thing is the mechanical and optical stability of the camera lens combo. A good magnesium body, like that of the Canon 40 D or the 5D by sambo3D, and a solid lens are very good for stable camera calibration, which is done on the fly in the aligning step by Photoscan. So a good choice, Bernhard! Some other photo hints: Don't change focal length or focus and turn off the image stabilizer.

Nevertheless, sometimes a good point-and-shoot camera with its bigger depth of field and good makro functionality (1-2cm distance) is a good alternative.

Cheers,

Karsten

How would the Canon EOS Rebel T2i 18 MP be as an alternative to the 40D? A lot less expensive, with much higher resolution. I am not a camera guy, so comparing the different models is difficult for me. I came up with the T2i after researching the Infinite Realities blog. Seems that they were using the European version of that model in their rig in a post from a year or so ago.

@scantastic

Basicly you can get good reconstructions with every DSLR as long as you have good light conditions. More resolution is helpful. I have change my Canon Rebel /300D to a Canon 60D last year and the models generated from my old camera were good too. The examples I have posted here from 2011 were made with the old camera.

Regards,

Karsten

Thanks for everything. I quoted few things I would like to better understand.

A_ Multi stereo approach, could you tell me a bit more about that?

B_ People eating around a table, yes, I would like to scan the people too. like in the example I am posting here (please if you can give me a feedback, low quality, 5000 faces). Why would I need more cameras than Infinite Realities? Apparently, I need more cameras than Lee as I need to build a rig around the whole table. That might be the answer if I wanted to produce a model suitable for very high end computer processing, or Tv broadcasting, or Video-gaming. But, what If I just need to have a rough (not abstract, just rough) version of my subject? I infer, from the whole discussion, that the only problem with using just one camera, and moving around the subject is the fact that the people might move. Right?
A more general question is: why does Lee need all those cameras?

Sorry for looking too noobish (that is the best term I was able to find to better describe me in this case).

C_Autodesk.... Oh my... Since when? they have something like 300K model already there! It is amazing, you just post pictures and they post-process them for free? the price is " you lose your (c) on the model)



thanks!
best
Alex

alternatively, try using a small aperture (11 - 16) and 100 iso. hence, use a tripod, I discovered how my little world turned upside down when I started to use a tripod! (or maybe downside up...)

firs three

last three

Hi Alex,



In short: the other name for SFM (structure from motion) is MVS (multi view stereo) which discribes the process better.
For example when you walk along an old stone wall and shoot every meter a photo perpendicular to the wall, so that the images overlapp to 80 % you get a first stereo image pair (photo A+B), a second (photo B+C) and so on. There is also an overlapp of 60% between photo A+C with a bigger base (photo A and C are 2 m appart from each other) and overlapp of 40% between photo A+D (3 m apart) and so on. From all these stereo pairs SFM/MVS software can calculate camera stations with all parameters, camera calibration and stereo models. Finally it generates one big 3D model. Of course the process isn't that easy just an imagination what is principly going on.

You can get a good overview at:
http://www.agisoft.ru/wiki/Photogrammetry



Lee is modeling characters in motion. Nobody can stand absolutly still without breathing. And he needs at least one (better more) stereo pair for every part of his character to be modeled. There is so much occlusion at humans (nose, chin, arms, legs, ...). You have tried a sculpture, you know how many positions are needed. When you have a group of people the problems are rising.



Yes they model it for free. There are other services you can try: http://www.hypr3d.com/ and http://my3dscanner.com/

Regards,

Karsten

Chào các bạn!
Tôi đã thử với photoscan và cảm thấy nó cũng tạm được với chi phí rẻ.



Tôi sử dụng máy ảnh FUJIFILM FinePix F800EXR. Chụp với độ phân giải 2MB và 24 bức ảnh


Tôi sẽ thử với chi phí lớn hơn bằng cách dùng phần mềm DAVID.


TRANS:
hi everyone!
I did try to photoscan, and it is also normal to feel cheaply.

I use the FUJIFILM FinePix F800EXR camera. Capture with 2MB resolution and 24 photos


I will try with greater cost using DAVID.