0.2.0a0

This is the first alpha release of birdvoxdetect v0.2
The main new feature is that birdvoxdetect now supports species classification.
The accuracy of the detector has also been improved with respect to the previous version.

0.1.2

This version is functionally identical to 0.1.0, with an updated README and relaxed Keras dependencies for PyPI distribution.

One important change is that the default detector does not use context adaptation (CA), but only a PCEN-based convolutional neural network: birdvoxdetect_pcen_cnn. To activate context adaptation on long recordings (longer than 30 minutes), set detector_name=pcen_cnn_adaptive-threshold-T1800.h5.gz.

The default value for the threshold is 30. Maximum sensitivity can be achieved by lowering threshold, even down to the value zero.

0.1.1rc2

This version is equivalent to the previous RC, with updated install_requires:

    install_requires=[
        'keras==2.0.0',
        'librosa==0.6.2',
        'numpy==1.13.1',
        'scipy==1.0.0',
        'h5py==2.6.0',
        'pandas==0.20.3'
    ],

keras has been pushed down to 2.0.0

0.1.1.rc5

This version updates Keras and TensorFlow to circumvent an issue in keras.load.load_model.
The list of dependencies is:

    install_requires=[
        'keras==2.2.2',
        'librosa==0.6.2',
        'numpy==1.15.2',
        'scipy==1.3.1',
        'SoundFile==0.9.0',
        'h5py==2.8.0',
        'pandas==0.23.4',
        'tensorflow==1.0.1'
    ]

Also note that the default detector is pcen_cnn, i.e. without context adaptation.

0.1.1.rc3

Updated install_requires again.

    install_requires=[
        'keras==2.0.6',
        'librosa==0.6.2',
        'numpy==1.13.1',
        'scipy==1.0.0',
        'SoundFile==0.9.0',
        'h5py==2.6.0',
        'pandas==0.20.3'
    ],

0.1.1rc0

This version is functionally identical to 0.1.0, with an updated README and relaxed Keras dependencies for PyPI distribution.

0.1.1

This version is functionally identical to 0.1.0, with an updated README and relaxed Keras dependencies for PyPI distribution.

One important change is that the default detector does not use context adaptation (CA), but only a PCEN-based convolutional neural network: birdvoxdetect_pcen_cnn. To activate context adaptation on long recordings (longer than 30 minutes), set detector_name=pcen_cnn_adaptive-threshold-T1800.h5.gz.

The default value for the threshold is 30. Maximum sensitivity can be achieved by lowering threshold, even down to the value zero.

0.1.1rc1

This version is equivalent to the previous RC, with updated install_requires:

install_requires=[
    'keras==2.0.6',
    'librosa==0.6.2',
    'numpy==1.13.1',
    'scipy==1.0.0',
    'PySoundFile==0.9.0',
    'h5py==2.6.0',
    'pandas==0.20.3'
],

0.1.1.rc4

This version is equivalent to the previous RC, with updated install_requires:

    install_requires=[
        'keras==2.0.6',
        'librosa==0.6.2',
        'numpy==1.13.1',
        'scipy==1.0.0',
        'SoundFile==0.9.0',
        'h5py==2.6.0',
        'pandas==0.20.3',
        'tensorflow==1.2.1'
    ],

Note that tensorflow==1.2.1 has been added.

0.1.0

First development version. This version is identical to the beta version.