Two scale in one model

[LiNo]

Hi, I am using Zephyr Lite and am having a problem with scale - two different scales in the same model from a consistent set of photo's, with consistent focal length. It is difficult to describe, but in the model linked to, the "island" section on the bottom left, is a different scale to the main body of the model. The island section is actually additional photo's of the central main section.

https://1drv.ms/u/s!AgdskftNUq2xi6Rh3i8fMGGfMSintw

Full disclosure: the photo's related to the island section and those from the main section were taken 24 hours apart. As a result, water trickling over the rocks is causing a different "wet mark" and so I am not surprised that they did not line up correctly. However, my question relates to the difference in scale between the two sections (you will need to look at the size / distance between marks on the rocks to see it - because of the scale differences, the size of the marks and the distance between them is different in the two parts of model). What causes scale differences like this and how can I avoid them?

I don't have a problem removing the offending photo's from the model and regenerating it. However, as the objects that I am looking to create models of are all outdoor, there are likely to be differences in light / etc. from one set to the next. I am trying to ensure that I don't have this problem with future data sets.

Regards,

Noel.

[Andrea Alessi]

Hello Noel,

it's hard to pinpoint exactly which photos are the problem as having only the zep allows me to only view the thumbnails. However, what has happened here is that probably you had a few images that "connected" wrongly two different pieces of the same area. The result is that you have two "misaligned" parts, and zephyr, trying to make the most sense out of it, does what it can. All starts from the camera orientation, and if some cameras are oriented wrongly, you'll get all kinds of artifacts. IN your case, the fix would be easy, by removing all those cameras "below" the ground.

Regardless, if you have to shoot in two different times of day, the correct process would be to shoot two different projects and merge them togheter with control points which, sadly, you can not do with zephyr lite. This is a very uncommon scenario for a lite user, but we are more than happy to help you salvage your dataset.

If you wish, i can send you a temporary key for Zephyr Pro so that you can finish your work. (the zep will be still compatible with zephyr lite afterwards).

[LiNo]

Thank you Andrea. Before I progress to merging projects, I want to try to get a good dataset(s) for all of the objects that I want to model.

In relation to camera orientation, naturally I did not have any cameras below the ground , but how do I orient a camera in the outdoors (I'm sure that it is obvious but I am using a camera and taking lots of photo's - should I do something to orientate the shots relative to each other)? Is there a section of the manual that deals with this and I'll read up on it further?

Regards,

Noel.

[Andrea Alessi]

hi Noel,

well, generally speaking you should simply keep overlap. If objects are moving during the day, then it's going to be a problem obviously.

If anything, i see big clusters where you are mostly still and just rotating - i would avoid that. too many images from the same spot may actually create noise / make things harder. Consider to always do a translation motion between two shots.

[LiNo]

Thanks Andrea. I'll try that.
By the way, I repeated the project with the same dataset on the new v4, it worked great (although it did take almost 24hours to complete - significantly longer overall than v3.7, but with much better results - not a problem, just letting you know).

Regards,

Noel.

[Andrea Alessi]

Hi Noel,

odd that it too so long, but happy to hear that 4.0 was able to process all the photos correctly

out of curiosity can you forward me the log, so i can take a look at the running times?

[LiNo]

Here you go Andrea. There is a "pause" of about 2 hours in it (at 15.36) which is why I mentioned "almost 24 hours". Again, not a problem (it just ran overnight).

https://1drv.ms/t/s!AgdskftNUq2xi7lFKzNHSJOKAjVy5g

Regards,

Noel.

[Andrea Alessi]

So i had a very quick look and i noticed that in some points unfortunately you ran out of memory, e.g.


[22:49:41] SysInfo | Cpu: 75.36 %; Disk (C/T): 288076/473815 MB (185739 MB free, 60.80 % used); Memory (C/T): 21950/16235 MB (2798 MB free, 135.20 % used)

[00:31:42] SysInfo | Cpu: 32.09 %; Disk (C/T): 290032/473815 MB (183783 MB free, 61.21 % used); Memory (C/T): 23795/16235 MB (727 MB free, 146.57 % used)

unfortunately performances hit a brick wall when this happens, as zephyr needs to swap data back and forth from memory to disk.

We are working to optimize this, as currently the only solutions you have when this happen are to either use lower parameters or increase the system hardware memory. You probably know this already, but make sure there are no other software running when using zephyr, as they sometimes may reserve system memory that zephyr won't be able to use. It's easy to forget, lets say, chrome with a thousand tabs open, in background, which may be eating some precious memory.

[LiNo]

Thanks Andrea. At these times, there should have been nothing else running on this machine but I will keep an eye on that. By the way, what is "Memory (C/T)"?

Regards,

Noel.

[Andrea Alessi]

You're welcome!

it's currently used vs total - so the system was currently using 21950MB out of the 16235MB available, for example.

[LiNo]

Understood. Thanks Andrea.

Regards,

Noel.