COLMAP crashes when using the feature extraction module

[Mathilde Marchandon]

Hello, 

I am trying to use the feature extraction module but COLMAP crashes when I run it. COLMAP is installed on a cluster under Linux (Debian). 

You can find below the message that I got on my terminal when COLMAP crashes.

Do you have any idea of where the problem could come from? 

Thanks a lot 

Mathilde

terminate called without an active exception

*** Aborted at 1554111307 (unix time) try "date -d @1554111307" if you are using GNU date ***

PC: @     0x7f5516e41fff gsignal

*** SIGABRT (@0x6c5fa00006647) received by PID 26183 (TID 0x7f551da95900) from PID 26183; stack trace: ***

    @     0x7f551c0580e0 (unknown)

    @     0x7f5516e41fff gsignal

    @     0x7f5516e4342a abort

    @     0x7f55179870ad __gnu_cxx::__verbose_terminate_handler()

    @     0x7f5517985066 (unknown)

    @     0x7f55179850b1 std::terminate()

    @     0x55eb1f2d69e1 colmap::internal::FeatureWriterThread::~FeatureWriterThread()

    @     0x55eb1f232ef1 colmap::SiftFeatureExtractor::~SiftFeatureExtractor()

    @     0x55eb1f233089 colmap::SiftFeatureExtractor::~SiftFeatureExtractor()

    @     0x55eb1f40fa88 _ZN9QtPrivate18QFunctorSlotObjectIZN6colmap19ThreadControlWidgetC4EP7QWidgetEUlvE_Li0ENS_4ListIJEEEvE4implEiPNS_15QSlotObjectBaseEP7QObjectPPvPb

    @     0x7f5517f22769 QObject::event()

    @     0x7f551c8ab30b QWidget::event()

    @     0x7f551c863b8c QApplicationPrivate::notify_helper()

    @     0x7f551c86b351 QApplication::notify()

    @     0x7f5517ef5d30 QCoreApplication::notifyInternal2()

    @     0x7f5517ef84cd QCoreApplicationPrivate::sendPostedEvents()

    @     0x7f5517f49dc3 (unknown)

    @     0x7f55109a37f7 g_main_context_dispatch

    @     0x7f55109a3a60 (unknown)

    @     0x7f55109a3b0c g_main_context_iteration

    @     0x7f5517f4a1cf QEventDispatcherGlib::processEvents()

    @     0x7f5517ef3d1a QEventLoop::exec()

    @     0x7f5517efc46c QCoreApplication::exec()

    @     0x55eb1f1ed044 RunGraphicalUserInterface()

    @     0x55eb1f1d6d43 main

    @     0x7f5516e2f2e1 __libc_start_main

    @     0x55eb1f1da9aa _start

    @                0x0 (unknown)

Aborted

[Johannes Schönberger]

Do you have a GPU in the system?

--
You received this message because you are subscribed to the Google Groups "COLMAP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to colmap+un...@googlegroups.com.
To post to this group, send email to col...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/colmap/1df95c3a-533d-4d2a-96f8-f4994bb5bc99%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Mathilde Marchandon]

No, is it why COLMAP crashes ? 

To unsubscribe from this group and stop receiving emails from it, send an email to col...@googlegroups.com.

[Johannes Schönberger]

You are trying to run the GUI, but you don't have a GPU. This cannot work. You have to stick to the command-line interface.

> >> To view this discussion on the web visit https://groups.google.com/d/msgid/colmap/1df95c3a-533d-4d2a-96f8-f4994bb5bc99%40googlegroups.com <https://groups.google.com/d/msgid/colmap/1df95c3a-533d-4d2a-96f8-f4994bb5bc99%40googlegroups.com?utm_medium=email&utm_source=footer>.

> >> For more options, visit https://groups.google.com/d/optout.
>

> --
> You received this message because you are subscribed to the Google
> Groups "COLMAP" group.
> To unsubscribe from this group and stop receiving emails from it, send

> an email to colmap+un...@googlegroups.com.

> To post to this group, send email to col...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/colmap/0d1b676d-b51e-4f42-829a-421c473f01a9%40googlegroups.com <https://groups.google.com/d/msgid/colmap/0d1b676d-b51e-4f42-829a-421c473f01a9%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Mathilde Marchandon]

Hello, 

Thank you for your answer.

I tried the feature-extractor tool using the following command line:

colmap feature_extractor --database_path /data/ist-calcul6/local/marchandon/landers/aerial1992/images/seg20/crop/testcolmap/database.db --image_path /data/ist-calcul6/local/marchandon/landers/aerial1992/images/seg20/crop/testcolmap/images  --ImageReader.single_camera 1

I obtained the following error message:

==============================================================================

Feature extraction

==============================================================================

Shader not supported by your hardware!

Framebuffer object not supported!

ERROR: SiftGPU not fully supported.

terminate called without an active exception

*** Aborted at 1554711287 (unix time) try "date -d @1554711287" if you are using GNU date ***

PC: @     0x7f9bd52eafff gsignal

*** SIGABRT (@0x6c5fa00009a71) received by PID 39537 (TID 0x7f9bdbf3e900) from PID 39537; stack trace: ***

    @     0x7f9bda5010e0 (unknown)

    @     0x7f9bd52eafff gsignal

    @     0x7f9bd52ec42a abort

    @     0x7f9bd5e300ad __gnu_cxx::__verbose_terminate_handler()

    @     0x7f9bd5e2e066 (unknown)

    @     0x7f9bd5e2e0b1 std::terminate()

    @     0x564d7887f9e1 colmap::internal::FeatureWriterThread::~FeatureWriterThread()

    @     0x564d787dbef1 colmap::SiftFeatureExtractor::~SiftFeatureExtractor()

    @     0x564d78795796 RunFeatureExtractor()

    @     0x564d7877fd43 main

    @     0x7f9bd52d82e1 __libc_start_main

    @     0x564d787839aa _start

    @                0x0 (unknown)

Aborted

What can I do to fix this issue ? 

Thanks a lot for you help

Mathilde

[Johannes Schönberger]

Seems like you don’t have a CUDA capable GPU. Try to use CPU based feature extraction/matching. Probably you want to redirect the number of features during extraction, otherwise matching will be very slow. 

To unsubscribe from this group and stop receiving emails from it, send an email to colmap+un...@googlegroups.com.

To post to this group, send email to col...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/colmap/9f697948-2c7f-4d53-9219-e5b0e332a9bf%40googlegroups.com.

[coas...@gmail.com]

Hello. I am as well having a problem with feature extraction. However, I have been having excellent success prior to a yet unidentified change that took place 4 days ago.

System:

NVIDIA GEForce GTX 1660

Windows10

Intel I7

16GB ram

I purposely had the card installed for the purpose of running COLMAP to perform human model reconstruction from photos

Was all working very well, and suddenly I began to get this error message:

"ERROR: Not enough GPU memory"

"ERROR: SiftGPU not fully supp"

Now I see that the GEforce GTX 1660 is not listed under CUDA-enabled GPU's list on the NVIDIA site

However I understand from my local technician resource that it is listed as such on the egpu website, although I can't find it there.

Perhaps there was a software update from NVIDIA which disabled the device?

It is indexed as -1 in COLMAP.

I do not know how to perform the C prompt programming however my local technician may be able to help me with this should you provide suggestions.

Thank you kindly for your assistance.

Best regards,

Roger

[coas...@gmail.com]

On second thought, also this could be caused by a Windows update issue?

r

[Pierre-Olivier Agliati]

Hi,

I don't think your card model is the problem, but it has "only" 6Gb of memory.

According to the size of your images, and the maximum features to be found, you may

have reached the limits of the GPU memory especially for the allocation of the matrix used

in the second step of feature matching. Have a look at the max_image_size,

max_num_features, and max_num_matches parameters in feature extraction 

and feature mapping steps. Lowering them may help.

Cheers,

Pierre-Olivier

--
You received this message because you are subscribed to the Google Groups "COLMAP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to colmap+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/colmap/09254c85-ca3e-4406-896d-2e5af3b44686%40googlegroups.com.

[Roger Rasmussen]

Hi Pierre,

Thank you and indeed you are correct on this. Late yesterday I found that reducing the max_num_matches value below 30,000 solved the issue.

The issue is solved for now, but I wish I understood the theory better. Perhaps with time and experience this will come.

Thank you,

Roger

To unsubscribe from this group and stop receiving emails from it, send an email to col...@googlegroups.com.

[jims...@hotmail.co.uk]