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.

Messages - jnb

Pages: 1 [2] 3 4
Indeed, exactly the same results, with and without masks...There is a real problem with the new depth maps.


And in the full release? (

I took the time to test again the v1.7.2 build 12070 and the mesh is still not good. I am still using the v1.6, the last bug free version for me.

edit : add exemple of issue on the depth maps and correct one on 1.6

Hello Alexey,

By any chance, did you have time to look further into this issue ?
I can confirm that such behaviour is present every time in 1.7.2 but with fluctuating strength.

Feature Requests / Add distance/gsd option to defocus mask generation
« on: February 23, 2021, 05:16:12 PM »

The defocus mask is a great addition to the 1.7 version, but I think there is still room for improvement. When dealing with close-range photogrammetry, some areas in the background could indeed still be in focus but lacking resolution in comparaison with the foreground, leading to visible differences in the orthomosaic. Adding a max distance and/or min gsd option on top of the defocus mask generation would be great in those cases.

Besides, could you detail how the defocus threshold works ? In some cases it seems to operate improperly, leading to in focus areas to be masked whereas out of focus ones are not.


Thanks for your response.
For this particular area, the 1.7.2 seems to work (medium quality, mild filtering) but the main issue related to holes/incorrect reconstruction in the mesh is still present as stated by Mak.
See the top of the model in attachment.

Feature Requests / Re: Iterative Closest Point Algorithm
« on: February 19, 2021, 01:47:57 AM »

I'm trying to use this script on v1.7.1. Thanks to the indications given by PolarNick the installation went fine but when running it I have the following error : AttributeError: module 'open3d' has no attribute 'registration'.

Some research seems to indicate that 'registration' has now moved to 'open3d.pipelines' but my skills in python are to limited to understand more...

Thanks for any help

EDIT : Thank you PolarNick for the updated script, works perfectly now !

Hello Alexey,

I have uploaded the data on your ftp and sent an email to support.
Thank you for your help

Hello Alexey,

Thank you for looking into it. I can send you the data but the images are big (37 Go folder). Would it be possible to have an ftp acces on your servers ?


Very good to know, as I also have this problem. Thank you for the fast resolve.
Are the errors in mesh reconstruction from depth maps as shown in attachement also linked to this specific issue ?

Thank you

General / Re: Assign images window keeps getting biger and biger
« on: December 23, 2020, 07:52:15 PM »
Thank you very much for the tweak, it works as intended. Hopefully it will be fix soon.

General / Re: Assign images window keeps getting biger and biger
« on: December 23, 2020, 03:14:48 PM »
I'm using the last version (not beta), but I had the same issue with every other versions.
I can confirm that when closing the dialog, opening it again the size will not be the same. Every time it gets bigger.

If it can help, here is my configuration :

- windows 10 64 bits
- i7 7820HK
- 64 Gb ram
- GTX 1070 SLI

General / Assign images window keeps getting biger and biger
« on: December 22, 2020, 06:00:33 PM »

I noticed some time ago a minor ui bug : when editing an orthoimage using the draw patch tool, the assign images window keeps getting biger and biger every time it is used. It therefore needs to be resized manually every time.

Not a big deal but a fix in the next version would be much appreciated because it's getting really annoying and time consuming when making a lot of ortho edit.

Thank you

General / Re: Photoassignment Orthophoto is bad
« on: September 14, 2020, 08:41:54 PM »

I second this, and not only for cylindrical projection bur for planar too.
As soon as I deal with a little bit complex orthomosaic the image assignment is awful, especially when multiple plans at different depths are involved.
It would be nice if we could manually select a metric threshold under witch images are not used during the process.
Here is an exemple before and after manual assignment. The part on the left is at a greater distance than the part on the right and as you can see the automatic orthomosaic uses the same images for both.


I need to move and resize the region according to specific coordinates, but after the application of a rotation matrix to the chunk.

So first, I apply the rotation matrix (see
Code: [Select]
chunk.transform.matrix = S * T
Then I use this script given by Alexey. The Metashape.vector values are picked in the new system (once the rotation matrix applied)
Code: [Select]
import Metashape

side_length = 0.97 #set value here
chunk =
T = chunk.transform.matrix = T.inv().mulp(Metashape.Vector([0.025, -0.1325, -1.0562]))
chunk.region.size = Metashape.Vector([side_length, side_length, side_length]) / chunk.transform.matrix.scale()
chunk.region.rot = T.rotation().t()

print("script finished.")

Unfortunately, this is not working as the region is not in the right place (wrong center), but I don't understand why.
If I use the same script with values for Metashape.vector coming from the original coordinates system and before the application of the rotation matrix, it is working as intended.

Any insights ?
Thank  you !

Python and Java API / [SOLVED] Apply transformation matrix to a chunk
« on: April 03, 2020, 10:17:10 PM »

I'm trying to apply a transformation matrix to a chunk in python, but I am struggling. 
My workflow for now :

- export a scaled model (local coordinates)
- open it in another software to give it correct orientation
- grab the corresponding matrix and apply it in Metashape

Exemple : 

Code: [Select]
import Metashape
doc =
chunk = doc.chunk
S = Metashape.Matrix().Diag([1, 1, 1, 1])

T = Metashape.Matrix([[0.9997743489568586, -0.0211954660930352, -0.0014154098306438, -13.4365763350189269],
                      [0.0000000000000000, -0.0666304947728083, 0.9977777193174494, 851.1973959656927491],
                      [-0.0212426732755011, -0.9975525697342621, -0.0666154595321578, -1634.9351050186285192],
                      [0.0000000000000000, 0.0000000000000000, 0.0000000000000000, 1.0000000000000000 ]])

chunk.transform.matrix = S * T.inv()

The model is indeed moving, but not correctly. The scale is also wrong (I have to hit update in Reference pane to get it right). So what I am doing wrong here ?
Thank you for your help !

Edit : ok, it seems I don't need to use the inverse matrix, chunk.transform.matrix = S * T is giving good results.

Pages: 1 [2] 3 4