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.

Topics - jrp

Pages: 1 [2]
General / Network cluster setup questions
« on: October 18, 2018, 02:00:04 PM »
I have 2 (moderately) powerful windows workstations and 2 headless Linux servers, I would like to set up a network cluster. There are some details not clear in the documentation though. Sorry for the long list of questions:

1)Can I run 2 workstations as both a processing nodes and a client nodes? To do this, do I need to run photoscan twice on each workstation (once from commandline with –node; and once from gui), or is there a way to configure photoscan to act as both node and client?

2)Computers on our site use windows domain logins, I notice many settings (including turning on GPU processing!) are stored as a per user setting, is there a simple way to configure things so that user training is minimal in how to add network processing on the workstations?

3)If running it as above, do I need 4 licences, or will the 2 licences I have be good enough?

4)I may consider buying more licences and running the 2 headless Linux boxes also as nodes (they are dual Xeon E5-2640 v3 with no graphics cards – so not ideal, but not slow either) – is running the server and a node on the same machine a problem?

5)Are there any reliability issues with running a mixed Linux windows environment? (other than the simple to fix --root file path stuff?)

6)We have a staff member with a laptop with photoscan on, can they easily connect up to the network and run as a client, then easily switch to standalone for field use?

7)Also, how much network traffic is there with the network processing? Would the system benefit from 10Gbit ethernet, or is gigabit good enough? How much does latency on the network file store affect performance? Do the nodes cache data locally? Would they benefit from SSDs?

General / Camera based chunk alignment - how does this work
« on: August 01, 2018, 07:25:23 PM »

I have been given a very large dataset to process, with only a very small number of GCPs (only one of my chunks has more than 1) and I need to somehow align the chunks.

I've just spent half a day visually searching for common points in more than one chunk, this doesn't seem to have helped with anything

Camera based alignment sounds like a great idea from the name, but I cannot find any documentation about how it works, or what I need to do to to get it to work. other than in the manual it says "Camera based
method is used to align chunks based on estimated camera locations. Corresponding cameras should
have the same label."

Presumably the same image file needs to be loaded into more than one chunk, and presumably "label" means file name?

Could really do with a decent youtube video of someone demoing this about now.

General / Sequence for combining Chunks
« on: May 18, 2016, 12:05:15 AM »
I am new to using Chunks.

I am processing details of architecture on the ceiling of a cathedral.

I have 2 chunks, one of a wider area including the walls of the room, and some ground control points, I added the images and ran camera alignment, added the GCPs then ran camera alignment, the results looked good.

I then added a second chunk, of a small area of the ceiling, aligned cameras, results for this chunk look good though not correctly oriented.

My intention is that the first chunk is only used to provide scale and orientation for the second.

When I align the chunks, the GCP errors become huge and the model ends up rotated wildly, losing all the effort of setting up the GCPs.

What am I doing wrong?

General / Bounding box rotation.
« on: December 29, 2015, 01:44:12 AM »
This may be covered elsewhere, but I suspect but part of my problem is terminology, so I don't really know what I'm searching for.

I have successfully got a local coordinate space set up and have cameras, scale bars and GCPs set up so that my building looks level when I press 1,2 or 3. This suggests that things are starting to work properly and that I'm starting to understand some of it.

I now have an arbitrarily lop sided "region" which I seem to have to rotate using a mouse until it surrounds the scene. questions:

What does it actually do?

Does it need to be level with respect to the model?

What are the effects of it not being level? I understand it has significant effects on height fields, though I'm not using height fields for this project.

How do I get it level (eyeing it in isn't good enough for me if it will affect the "levelness" of the finished product)?

The only way I can see to control its rotation numerically seems to be via Python, suggesting that either it doesn't matter (not likely) or I've missed something.

"Reset Region" appears to set it back to it's original lopsided, and almost perfectly upside down state.


Feature Requests / GCPs with relitive/wildcard coords
« on: December 23, 2015, 10:48:19 PM »
Related to

Sometimes a GCP position is only known as a position relative to another GCP and sometimes without a full 3 coordinates known.

examples could include:

2 arbitrary points with a plumb bob or other hanging string between them;

intersection of an arbitrary shaped pool of water with objects;

2 arbitrary points known to be a certain compass bearing from each other (but at different unknown altitudes);

a point known to be a certain distance above the floor of a room, but it's position otherwise unknown.

Scale bar support is a special case of the above. GCP is maybe the wrong concept here, but you get the idea.

Note that entering 2 GCPs with fully fixed locations does not work at all, and has no effect on the location or orientation of the model -- this may actually be a bug -- see link for more details.

General / Getting vertical right
« on: December 22, 2015, 02:51:36 AM »
I have asked around and scoured the internet, and am not finding what I need.

While scanning the inside of buildings (churches), I struggle to get the building the right way up. It seems that the only way to do this is to eye it in using the rotate tools in photoscan. Given that the buildings are old and have no single straight edge in them, this is fundamentally not a good a way to go.

I expect the official answer is the provide 10+ ground control points -- but I simply have no way of getting ground control points. I could get a total station, but that would defeat the point of using Photoscan.

I do not need a fully GCP located scan, all I need is vertical correct and the scale right. Scale is easy. I do not mind where north is.

I tried using 2 targets and a plumb bob to get them precisely located above one another, then measured the distance between them. This should be enough information to get the correct answer. I entered the first one as coordinates 0,0,0 and the second as the correct relative location. No other GCPs were entered, camera alignment was then done, and photoscan still insists on creating the whole building on it's ear, and complains that the GCPs are ~15 meters from where they should be.

I have just (moments ago) entered a third GCP by guessing it's location, and things have now changed in that the errors have now dropped down to something vaguely sensible, but the building is still lop sided as the guessing wasn't very good.

It seems this means Photoscan is ignoring GCPs unless there are at least 3 of them??

I could try adding a 3rd target right next to one of the others, but that doesn't seem like a good solution.

Is this a feature request?

Either way, this project is borderline unfinishable right now, despite the fact that I have all the data theoretically that I need.

Any advice on how to proceed will be gratefully received.

I'm using 1.04 build 1847 on a mac.

Pages: 1 [2]