Segmentation Fault when conducting change detection tutorial

[Charliechen1]

Hi,

Everything seems ok for the feature extraction tutorial. But when I train the model following exactly what the tutorial asks me to do for change detection, I got segmentation fault. What might be probably the reason. Thank you for your help.

[hedonistrh]

Hi,

I am not the owner of repository. Can you share your exact error with codes and error message?

[Charliechen1]

Seems the project failed in catching this error. It's just 'segmentation fault' in C. No other information is provided..

[chienducnguyen]

@Charliechen1 can you tell me how i can find "~/.pyannote/db.yml" or i need to create a db.yml

[hedonistrh]

@chienducnguyen If you work on Jupyter Notebook, you can do like this.

To create, your own db.yml, you can use any text editor. We should write the directory of AMI Corpus. In this case, it is AMI: ../pyannote-audio/amicorpus/*/audio/{uri}.wav

I want to clarify more.

In that case, I create folder named as .pyannote in the same folder with pyannote-audio. (Not in the pyannote-audio folder.) After that, I create file named as db.yml with gedit or VS code and write the AMI: ../pyannote-audio/amicorpus/*/audio/{uri}.wav In the end, you need to save this db.yml into .pyannote folder.

After that, you can run this command.

! cat ./.pyannote/db.yml | grep AMI

Also, you can check this.

[chienducnguyen]

@hedonistrh Tks for your help!
Now i'm stuck with
pyannote.audio.features.utils.PyannoteFeatureExtractionError: No precomputed features for "AMI/EN2002b.Mix-Headset"
when i run this command
$ pyannote-speech-feature ${EXPERIMENT_DIR} AMI.SpeakerDiarization.MixHeadset

[hedonistrh]

@chienducnguyen Can you check your data to understand you have or not EN2002b.Mix-Headset.wav :)

[chienducnguyen]

i already have EN2002b.Mix-Headset.wav and still got that issue :(

[Charliechen1]

You would like to download all the audios.

…

Sent from my iPhone

On Aug 11, 2018, at 11:16 AM, Chiến Đức Nguyễn ***@***.***> wrote: i already have EN2002b.Mix-Headset.wav and still got that issue :( — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[chienducnguyen]

it still need EN2002b although i download it :/

[chienducnguyen]

now it got that issue when i run extraction
$ export EXPERIMENT_DIR=tutorials/feature-extraction
$ pyannote-speech-feature ${EXPERIMENT_DIR} AMI.SpeakerDiarization.MixHeadset

AttributeError: 'Precomputed' object has no attribute 'dimensions_'

[hedonistrh]

@chienducnguyen May you check the output of this command in the your terminal?

pip show pyannote.audio
You should get the output which start with these information, if you use the latest version.

Name: pyannote.audio
Version: 1.0.1

[chienducnguyen]

It show the same with you

[hedonistrh]

If you work on Jupyter Notebook, can you execute this code instead of yours.

!pyannote-speech-feature tutorials/feature-extraction AMI.SpeakerDiarization.MixHeadset

[hbredin]

Closing because the issue has diverged from the original issue (segmentation fault).
Please open a new issue for every new problem...

@Charliechen1 Segmentation faults have happened to me before. The culprit was dlib which was not installed correctly. If import dlib raises a segmentation fault, I suggest you try fixing this first...

[wuzl]

hi @hbredin, how to fix the segmentation faults error. import pyannote.audio.train return Segmentation fault

[hbredin]

Try import dlib. If you get a Segmentation fault, the culprit is likely to be dlib.
You're going to have to fix your dlib installation.