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 - tutoss

Pages: [1] 2 3 4
General / Re: Negligible GPU usage on Alignmnet.
« on: August 15, 2020, 10:07:54 PM »
Well, 50mp are pretty big hehe, plus they are 16bit.
about the two cameras, in this step (detecting points) is irrelevant.
If you wish, you can send me a small dataset (~200 images maybe) to benchmark the times with and without gpu (remember we have "similar" cpu and gpu).



General / Re: Negligible GPU usage on Alignmnet.
« on: August 14, 2020, 08:35:57 PM »
Thank you guys, this is helpful.

This particular block is comprised of ~4000 medium format images as an added bit of info.

In the preferences>advanced>tweaks section there is an option, "main/refine_max_gpu_multiplier" which has a value of 2. I cannot find any documentation on this anywhere. Any ideas as to what it does?

Which is your photos resolution?
My times for "detecting points" are with 21MP photos (5472x3648), Phantom 4 rtk.
I tested with a similar project (3500 images), with one 2080 Super, and the time to finish "detecting points" is ~27 minutes.
Without GPU, the time up to ~1 hour 50 minutes.

We have similar CPU and GPU, and you says your time is 6 hours for 4000 images!!! or the image resolution is very very high or something more is happening.

About "refine_max_gpu_multiplier" i don't know, the only thing i change on metashape are enable all GPU and turn off the CPU for GPU usage.



General / Re: Negligible GPU usage on Alignmnet.
« on: August 13, 2020, 11:17:14 PM »
Also "low" gpu usage (see attached image) on two 2080 super and one 1070.
Anyway, 7 minutes for "detecting points" for 843 photos, is very good! 

I tested with all gpu disabled, and the time for run "detecting points" is about 32 minutes (AMD TR 2950X).
With one 2080 super, the time is 7,5 minutes (the true power of 3 gpu is much more noticeable on other stages).

Metashape 1.6.3

You can try benchmarking with all gpu disabled and share the results.



General / Re: Markers with fewer than three coordinates (x,y,z)
« on: April 13, 2018, 09:34:47 PM »
wow, that is useful, it will be in the GUI, no "hidden"

General / How is AMD Ryzen performance on photoscan v1.3?
« on: April 19, 2017, 07:12:20 PM »

In a Anandtech review (, Photoscan v1.0 got mixed results vs Intel. Since this benchmark in based on on the old v1.0, i was wondering how it's perform with all the tweaks on the lastest version.

Any info?



General / Re: Why does my model look so miserabel?
« on: November 10, 2016, 02:25:50 AM »
Hi hermiew,
well, to use markers, you need coordinates of them.
Try take more photos and align them without markers

General / Re: Banana shape effect
« on: September 26, 2016, 06:08:14 PM »
you can try add a previous camera calibration parameter and fix it. Works for me.


Feature Requests / Re: Rolling shutter integration
« on: September 21, 2016, 07:51:56 PM »
any news on this?

Feature Requests / Re: Rolling shutter integration
« on: September 11, 2016, 07:54:59 PM »

Feature Requests / Re: Rolling shutter integration
« on: August 03, 2016, 11:00:15 PM »

Feature Requests / Re: Show progress in taskbar
« on: June 22, 2016, 12:01:47 AM »

there is uploading via wetransfer, splitted in two files.

Chunk properties:

Hi Alexey,
Yes, is the 64bit version (checked in about Photoscan...)

Wetransfer link of the log:

settings used (same on 2014 (1.0 i think) and 2015 with Photoscan 1.2.1 due to expansion of the project (more images)):
accuracy: high
pair preselection: reference
key point limit: 40000
tie point limit: 0

Precalibrated camera with same file on both years

The first error is with all images referenced but with the elevations forced to zero
The second error is with all images without Z modification (original process)
The third error is the abort of the current process (half of images)



General / error: bad allocation (not enough memory) only on 1.2 (1.2.3)
« on: January 28, 2016, 05:03:03 PM »
i got this error with this version in the alignment step. Is a big old Project (8600 images) and i reprocessing it for test purposes.
With 1.0 and 1.1 is processed without issues in the same pc (64 gb ram) and i processed projects with 21000 images on v1.1 without problems  :-\.

General / Re: Poor quality results in 1.1.6 build 2038 (64 bit)
« on: December 18, 2015, 03:53:55 AM »
Hi alexey,
the tipical settings for aerial data with geolocation info:

accuracy: high
pair preselection: reference
key point limit: 40000
tie point limit: 0 (to replicate Photoscan 1.0)

Dense cloud medium and aggresive filter.

Pages: [1] 2 3 4