RuntimeError: Can't decompress depth

[EE]

We're processing 2 completely different projects using the python API. In 1.6.3 this worked, but in 1.7.4 and 1.7.5 these seem to fail with the following code and logs:

Code: [Select]

openexr error: Error reading pixel data from image file "openexr". Unexpected tile x coordinate.
openexr error: Error reading pixel data from image file "openexr". Unexpected tile x coordinate.
openexr error: Error reading pixel data from image file "openexr". Unexpected tile x coordinate.
openexr error: Error reading pixel data from image file "openexr". Unexpected tile x coordinate.
Traceback (most recent call last):
chunk.buildDenseCloud(point_colors=True,progress=XXX)
RuntimeError: Can't decompress depth


[Alexey Pasumansky]

Hello EE,

In most cases this message indicates to the issues with RAM which led to the corrupted data in the depth archives. Therefore, if you observe the problem on the same computer, try to run mutithreaded Memtest86 checks overnight to see, if any failures are detected.

If the system is overclocked, try to disable the overclocking.

[wizprod]

I have gotten the same error in 1.7.4 in an almost brand new system with 128 gb of ram. I have run mem-test, and there is nothing wrong with the memory.

Additionally, I am using normal jpegs in my project, so why is this failing in openexr?

[Alexey Pasumansky]

Hello wizprod,

the depth maps data is stored internally in exr format. The error appears, when the depth data is required and archives with the depth maps are unpacked.

Have you started Memtest86 in multi-threaded mode? You may also try to lower the RAM frequency via BIOS and check if it helps.

[EE]

Hey Alexey,

We observe the same failure on 2 different laptops so the chance it's related to RAM is quite low.
Not overclocked btw.

Greetings

Evert

[Alexey Pasumansky]

Hello Evert,

Does the issue appears on both test systems, if you generate depth maps on each one of them directly?

Re-using the corrupted depth maps data for build dense cloud from the older Metashape project cannot be fixed, unfortunately. As the lost data cannot be recovered without re-calculating the depth maps.

[EE]

Yes both start a project from scratch.