pyannote/deprecated-pyannote-db-template

[deprecated] Template for creating your own pyannote.database plugin

deprecated-pyannote-db-template's Issues

Error while testing the edited template

I noticed that Tensorflow installation is not mentioned in this tutorial. Following the instructions I get an error:

ImportError: No module named 'tensorflow'

I installed it now with pip install tensorflow.

I have modified the template like this, and I try to do the training with:

export EXPERIMENT_DIR=experiment
pyannote-speech-detection train ${EXPERIMENT_DIR} ikdp.SpeakerDiarization.MyFirstProtocol

The error I'm getting is right in the beginning of the first epoch, here is the complete output:

Using TensorFlow backend.
/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/database/util.py:193: UserWarning: "annotated" was approximated by "wav" duration.
  warnings.warn('"annotated" was approximated by "wav" duration.')
2017-07-13 16:43:05.736378: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-13 16:43:05.736413: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
2017-07-13 16:43:05.736420: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-13 16:43:05.736427: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.
Exception in thread Thread-1:
Traceback (most recent call last):
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/threading.py", line 914, in _bootstrap_inner
    self.run()
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/threading.py", line 862, in run
    self._target(*self._args, **self._kwargs)
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/keras/engine/training.py", line 429, in data_generator_task
    generator_output = next(self._generator)
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/generators/batch.py", line 416, in __call__
    raise e
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/generators/batch.py", line 409, in __call__
    current_file, identifier=uri)
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/audio/generators/speech.py", line 86, in preprocess
    coverage = annotation.get_timeline().coverage()
AttributeError: 'Timeline' object has no attribute 'coverage'

Epoch 1/1000
Traceback (most recent call last):
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/bin/pyannote-speech-detection", line 11, in <module>
    sys.exit(main())
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/audio/applications/speech_detection.py", line 522, in main
    application.train(protocol_name, subset=subset)
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/audio/applications/speech_detection.py", line 297, in train
    optimizer=SSMORMS3(), log_dir=train_dir)
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/pyannote/audio/labeling/base.py", line 153, in fit
    verbose=1, callbacks=callbacks)
  File "/Users/niko/anaconda3/envs/py35-pyannote-audio/lib/python3.5/site-packages/keras/engine/training.py", line 1532, in fit_generator
    str(generator_output))
ValueError: output of generator should be a tuple (x, y, sample_weight) or (x, y). Found: None

There are some variations of the error message depending on what I try, but invariably present error message is always that on the last line.

I didn't put my mdtm file yet to GitHub as I may want to anonymize the identifiable info that is in the speaker id's first, but it looks basically like example below, I was also thinking the problem could be related to it somehow. For now I selected recordings which are completely annotated:

file20090222 1 0.0 1.663 speaker NA unknown aaa
file20090222 1 1.663 2.289 speaker NA unknown aaa
file20090222 1 3.952 1.256 speaker NA unknown aaa
file20150402-7-b 1 2005.1 4.206 speaker NA unknown bbb
file20150402-7-b 1 2009.261 2.112 speaker NA unknown ccc
file20150402-7-b 1 2011.32 0.52 speaker NA unknown bbb

It is entirely possible that I have just missed some clear instruction somewhere, I appreciate if pointed to the right direction if that's the case. Thank you! Merci!