Messages

46

Face and Body Scanning / Re: Smart Shooter, Nikon D3200 and noiseprojection.

« on: July 01, 2013, 06:07:28 AM »

thanks Magnus! yes we are having that problem with trigger delay. appreciate the info!

47

Face and Body Scanning / Re: Smart Shooter, Nikon D3200 and noiseprojection.

« on: June 29, 2013, 10:37:45 PM »

thanks for sharing your info Magnus!
hi just a quick question, why not trigger with smart shooter instead of remote?  i'd like to be able to software trigger canon and nikon dslrs.

48

Other Languages / Re: Маркеры

« on: June 01, 2013, 09:39:50 PM »

Спасиба Алексей,

Да, я про Full Body capture говорю, иногда нет возможности выбирать одежду. При таких ситуациях становится сложновато.

Я думал если есть какието специфики алгоритма которые можно использывать в таких ситуациях.


Руслан

49

Other Languages / Маркеры

« on: May 18, 2013, 09:13:46 PM »

Есть маленький вапрос по маркерам.

Имея только Стандартную версию PhotoScan, можно ли както использывать реальные маркеры или фигурки на одежде человека, чтобы помочь с Alignment-ом?

Я понимаю для етого есть Pro версия, но хотелось бы както понять, что еше можит помочь с реконструкцией. Я не пробывал ето еше, но есть такое ошущение что PhotoScan-овский алгоритм в некоторых случаях можит "перевешивать", то всякие маркеры.

Еше есть идея с Grid-ом нарисованым на платформе где будит стоють человек. Есть ли в нем смысл?


Большое спасибо,
Руслан

ПС. Заранее извиняюсь за мое написание, не часто на русском пишу.

50

Other Languages / Re: Контроль над автоматическим состыковыванием чанков

« on: April 23, 2013, 06:02:00 AM »

Spasiba balshoe Alexey.

51

Other Languages / Контроль над автоматическим состыковыванием чанков

« on: April 17, 2013, 11:10:51 PM »

Поевился вопрос один по поводу автоматического состыковывания чанков.

У нас есть сет-ап из 20+ камер, расположеные так что аррей захватывает примерно треть 360й ротации, тоесть 120 градусов. Первым шагом снимается фронтовая часть человека, тоесть лицо, и потом по поровороту по каждой оставшейся стороне.
 
Но иногда, когда мы делаем реконструкцию человека так случается что он нимного улыбается на одном из кадров. И получается так что две фотки с нимного разных углов реконструируются одна по верх другой, что нарушает качество реконструкции.

Вобщем вапрос таков, можно ли за-маскать видимые погрешности в таких случаях какимто способом? Патаму как не очень хочется убивать целую фотку изза того что только скажем 10-20% етой фотки несовместивы?


Спасибо,
Руслан

52

Face and Body Scanning / Re: 4D Scanning?

« on: February 18, 2013, 10:38:59 PM »

Anyone tried with the new GoPro Hero 3 ?

4K at 15fps might be ok and the price reasonable. You can also sync them with the remote (up to 50).

I've tried the GoPro Hero3 as soon as it came out, but the results were not impressive.
The wide angle lens requires the subject to be way too close to the camera, otherways there's a major los of detail (and 12mp in not much anyway). So there's really no room for action.

The optics are pretty subpar, the claimed 4k resolution is a rescaled 2.7k. Still lots of noise, aberration, autoleveling: the whole package.

53

General / Re: Unexpected application of Pscan

« on: February 01, 2013, 12:06:47 AM »

Also, don't forget that George mentioned that he converted his pointcloud to a triangulated mesh using ArcScene before importing it int PScan.

54

General / Re: Quality Improvements

« on: September 13, 2012, 08:13:10 AM »

Hello... as far can be seen it is reconstructed on MEDIUM - HIGH, not ULTRA settings. then is only way to mask it. If you use Pshop ( CS5- CS6 ) ther is a very ussefull tool, press letter W and it shold be activated QUICK SELECTION TOOL ( not the MAGIC WAND TOOL !! ) the 4th tool from TOOL palette, move cursor with pressed LEFT mouse button over areas what need be masked, dont forget to INVERT mask.... Fastest it goes with custom shortcuts..

The problem is that ULTRA calculate every pixel, but High just every 2nd and Medium every 4th pixel of image...... Like you mask go outside of object and the pixels are counted..

Try it and let us know if that helped...

Thanks Wishgranter,

Yeah, i did have a hunch it was because of the UltraHigh settings, but still, my expectation was that this oversampling would be avoided by getting rid of any extra points floating around.
Would be nice to see an improvement in this area maybe in the next update, just so the results are a little bit more predictable, cause it does take nights to see these results.

Hello Rus,

Looks like the photos of the object are made only from one side, so Smooth reconstruction method tries to reconstruct closed model and therefore additional geometry appears behind the object.
In general there could be three main options: reconstructing in Sharp (with no additional geometry), using masks for background, or taking photos from all sides of the object so PhotoScan could create closed model.

Please also note that geometry is reconstructed based on camera positions and not only by points in Sparse point cloud. So to refine camera positions Optimization should be run after some points are removed (currently only in Pro version).

Meaning of P1 and P2 parameters are described in the corresponding topic here: http://www.agisoft.ru/forum/index.php?topic=19.msg45#msg45
But in your case changing these parameters is not a solution.

Thanks for the info Alexey,

I was a little too lazy and didn't capture all sides of the subject, since it was more or less a test. But, sometimes it is hard to get all the details or angles in a non-constrained environment. Maybe there could be an option to tell PS a degree of tolerance for these puffy, undersampled areas?

And.. by the way, I did look at that P1P2 post previously, but it still did not explain much. :S
I mean like, what units these parameters are in?

55

General / Quality Improvements

« on: September 12, 2012, 11:37:07 AM »

Hey guys,

I was wondering if there's any way to reduce blobby-ness without using masks.

Seems like this extra volume still appears even if I delete "all" extra points outside and inside the bounding box area after the alignment process.

That's actually one of the things that really throws me off, when everything looks clean and then boom after the heaviest and the longest process I get something, which can sometimes look completely different.

Or, let me put it another way, my expectation of the resulting mesh would so that the Geo conforms closer to the pre-selected PointCloud rather than the bounding box volume the mesh is in. Alot of times I feel like I'd rather have less than more.

I also need a little more explanation, if you can of what these P1, P2 parameters do in Preferences dialog. Even if its hard to explain just throw a few keywords at me that are a little bit less abstract.

Any hints? tips, tricks? ...money?


Cheers,
Rus

56

Python and Java API / Re: command line parameters

« on: September 05, 2012, 02:51:33 AM »

Or better yet, run a progressive loading of images sourced from a location on a disk with an automatic output of the latest geo.