Proces hangs during 'Estimating surface', can not cancel to save depth maps

[Casper_H]

Hi,

I'm  processing a city maquette in 4 chunks, but in the 'Estimating surface' stage of the first chunk, the progress just stops. The depth maps that were generated in over 10 hours cannot be saved, because the process cannot be cancelled; in taskmanager it shows that the CPU is fully utilizing a single core but the process does not get cancelled.

Is there a way to save the depth maps - outside of the GUI - so i can reuse those for another try?

Nvidia driver: 431.7 - Studio Driver
Windows: 1903
Metashape: 1.5.4 build 8885

With kind regards,
Casper

[wojtek]

The depth maps are saved in the .files folder regardless of keep depth maps option. You can manually re-add it to your project xml files and re use them.

That tweak should take 5 to 10 minutes.

[Casper_H]

Hi wojtek, thanks for your answer!

I have actually once tried that, but after opening the file, the depth maps did not show in the Metashape gui. I must've done something wrong, i'll try it again and post the results somewhere tomorrow.

Cheers, Casper

[Casper_H]

Hi again,

I'm starting to think that all is lost for these depth maps.

When i look at other (completed) project files, the pyramid folders are empty and there is 1 depth maps folder. In this project the pyramid folder contains a ~33,5 GB pyramid file.

It worked, i overlooked a change in one of the xml's because you should never open an xml with IE!!!

____________________________________________
Changes i made:
I have edited the xml from the chunk.zip and added the following:

Change 1: after -> </camera>
<depth_map_sets next_id="1" active_id="0">
    <depth_maps id="0"/>
  </depth_map_sets>

Change 2:
In the depth maps folder, i added the block paths of the depth maps as in other finished projects.

Change 3:
This is what i failed to find the first time:
Change in the .xml in the frame.zip file. (When you open and search in *cough* Internet explorer, it often does not find what you're looking for because it does not immediately open the whole file)
added:
<depth_maps id="0" path="depth_maps/depth_maps.zip"/>

Change 3:
I removed the depth_maps folder (which only had 2 files) and renamed the large depth_maps.1 folder to depth_maps.
_________________________________________

I have now only 2 important questions:

1) What about the large pyramid file. Do i still need it if i only want to salvage the depth maps and use them to redo the step 'Build mesh' -> ' Reuse Depth maps', or can i delete the file for now?

2) I removed the smaller depth_maps folder; can this folder be of importance for the other depth_maps folder? Or are these separate and can i delete them?



Casper

[wojtek]

Hey,

This is exactly what i meant, good job

removing depth_maps folder should be fine, that is another instance of depth maps and the bigger depth_maps.1 was what you were after. You can always also look at date modified values in explorer to confirm.

I think other files can go, unfortunately further stages of processing are not reusable. In my opinion it should be possible as it is saved out in chunks/blocks as well but metashape prefers to start over

[Casper_H]

Alright, i will first do the other chunks and then try to redo this recovered chunk.

And absolutely, it would be a great help if the 'option save file after each step' (in Batch process section) would also save after the depth maps were generated. I know it should not be necessary since the process should not crash in the first place, but currently it would be a great pragmatic, and easy to implement feature/change.

Thanks again for the tip, wojtek.

 

[Casper_H]

Although i have saved the depth maps, the initial problem persists:  application hangs at the same exact point every time (at 7% into 'Estimating Surface').

I have updated to the latest MS (1.5.5).

Everything is up to date, my RAM is only utilized at 40% or so (have 64GB RAM and around 500 GB of virtual mem).

[Casper_H]

Found the problem i think; i'm pretty sure the problem occurs when setting the desired polycount to more than the default value of 9.000.000.

I had set it to 20.000.000, and also once 12.500.000 i believe, and every time i tried this, the process hanged.

Does anyone have the same experience? It's pretty important to know i think, since it's often desirable to have  > 9.000.000 polys. Still, i'm happy i can go forward with this now.

[Alexey Pasumansky]

Hello Casper,

Please send the complete processing log, screenshot of the Build Mesh dialog with the parameters used and PDF report related to the source project to support@agisoft.com for more detailed investigation.