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 - Arie

Pages: [1]
General / 1.7.0. Results of reworked depth maps generation algorithm
« on: January 07, 2021, 11:42:26 PM »
Hi Agisoft team,
congratulations on the 1.7.0 release. I've been doing some comparisions between 1.6.5 and 1.7.0 and I thought I would share the results.

Agisoft 1.6.5: 12h 8m (Depth maps 9h 57m, Meshing 2h 11m)
Agisoft 1.7.0: 7h 14m (Depth maps 5h 9m, Meshing 2h 5m)
Competitor: 5h 11m (Depth maps 3h 5m, Meshing 2h 6m)

In general the improvements are quite impressive: Significantly shorter processing times with more detailed results. In addition, the stepping effect visible in some of the 1.6.5 images vanished.
But some areas of the 1.7.0 model show artefacts that were not present in 1.6.5 (last image). Any fixes in sight?

Also I am very impressed on how much Metashape has caught up to their main competitor. Keep it going! <3

Hi all,
actually this is not a bug but more of a heads-up to fellow Metashape users.

I'm currently running some tests to compare different settings. During my test, I noticed a significant boost in performance when checking "Use CPU when performing GPU accelerated processing" even though it says "When using dedicated GPUs please turn off ... CPU for optimal perfomance".
Here are the times (Metashape 1.6.5):

Unchecked:          6 hours 25 min (Depth maps), 32min (Build model)
Checked:               4 hours 9 min (Depth maps), 32min (Build model)
Unchecked 1.7:  1 hours 56 min (Depth maps), 35 min (Build model)
Checked 1.7:       2 hours 8 min (Depth maps) , 32 min (Build model)
Additional information: 269x 36mp images of a animal skull (i.e. turntable style data acquisition). Depthmap-based reconstruction in ultra high quality.
System: Threadripper 3960X, 256GB 3200 CL16, 1080ti, Samsung Evo Plus 970 Raid0 2TB.

My guess for the significant increase in performance is the high CPU core count of the 3960X. So it might be only helpful when using an HEDT system.
What I also find very interesting is the additional performance improvement of Speed-wise this is on-par with competing software. I'll be posting some comparisions in the 1.7 thread.
Hope this helps someone.

Dear Agisoft team,
is it possible to tweak the settings to change the resolution of the images used for generic preselection?
Being able to adjust this setting would let the user decide the balance between accuracy and processing time.

Bug Reports / - Private messages broken (?)
« on: September 15, 2020, 12:42:11 PM »
Hi all,
in the "My Messages" section, the contents of received messages are overlapping each other, making it impossible to read. Tried on chrome and firefox, both exhibit the same behaviour.
Is anyone else experiencing this?

Python and Java API / Aligning chunk based on 4x4 matrix
« on: January 27, 2016, 04:32:28 PM »
is it possible to align a chunk (translation + rotation) based on a 4x4 matrix?

I have a couple of datasets, that have been referenced by overlapping areas using an ICP algorithm (done in a different software). I have the 4x4 matrix of the transformation and would like to use that for referencing the chunk within Photoscan for DEM creation etc.


Feature Requests / Refining generic alignment
« on: January 18, 2016, 05:04:09 PM »
using the generic option for alignment has some major speed advantages but I've had quite a lot of situations where it produces fairly inaccurate results in comparision to the "disabled" align option.

Would it be possible to use the generic alignment as the basis for a finer alignment? For example using the camera-coordinates to run a second alignment using the "reference" option?

Edit: Or would it be possible to expose the parameters for the generic alignment?
This could be a trade-off between the higher accuracy of "disabled" alignment and the speed of "generic" alignment.

Bug Reports / Error reading file
« on: December 08, 2015, 03:27:28 PM »
Dear Agisoft team,
i've just finished a project and after saving it I now get a "Error reading project". In the console it states "Error: disk io error".
I've tried to repair the zip files in the .files folders, but some of them can't be repaired.
Any other chance to recover the dataset? Multiple days of calculation would be lost...

Further info: Photoscan 1.2.0. Build 2181 Windows 7 Pro 64bit
Aligned, Dense Cloud, Textured Mesh.


Pages: [1]