Vesper

This release:

* Adds a `Help` dropdown that includes a link to draft [Vesper documentation](https://vesper.readthedocs.io/en/latest).
* Offers a reorganized navbar with clarified command names.
* Improves instructions on command forms.
* Improves clip calendar and clip album error handling.
* Adds a `vesper_admin createarchive` command that simplifies new archive creation.
* Allows control of which clips data are exported for in `Export clip metadata to CSV file` and `Export clips to audio files` commands.
* Adds support for [BirdVoxDetect](https://github.com/BirdVox/birdvoxdetect) 0.2.x, including both adaptive threshold and fixed threshold versions of the detector.

To install Vesper 0.4.8 in a Conda environment named `vesper-0.4.8`:

conda create -n vesper-0.4.8 python=3.6
conda activate vesper-0.4.8
pip install numpy==1.16.4
pip install vesper==0.4.8

[](https://doi.org/10.5281/zenodo.3708420)

Release Candidate 0 for Vesper 0.4.8.

This release:

* Adds two new NFC detectors, MPG Ranch Thrush Detector 1.0 and MPG Ranch Tseep Detector 1.0.

* Adds an optional detector score to clip album clip view labels. BirdVoxDetect and the new MPG Ranch detectors produce detector scores, but the Old Bird Redux detectors currently do not.

* Makes the Transfer Call Classifications command visible by default in the Vesper UI.

* Adds a classification argument to the Export Clips CSV File and Export Clip Audio Files commands, fixing issues 123 and 124.

* Uses Ruamel YAML instead of PyYaml, fixing issues 97 and 154.

To install Vesper 0.4.8rc0 in a Conda environment named `vesper-0.4.8rc0`:

1. Create the new Conda environment and install some Vesper dependencies in it with:

conda create -n vesper-0.4.8rc0 -c defaults -c conda-forge django h5py jsonschema librosa=0.6.3 pandas pyephem python=3.6 ruamel_yaml scikit-learn sqlite sqlparse

2. Activate the environment with:

conda activate vesper-0.4.8rc0

3. Install TensorFlow and Keras with:

pip install tensorflow~=1.12.2 keras==2.2.4

4. Install Vesper with:

pip install --extra-index-url https://test.pypi.org/simple/ vesper==0.4.8rc0

[](https://doi.org/10.5281/zenodo.3382243)

This release:

* Adds the [BirdVoxDetect](https://github.com/BirdVox/birdvoxdetect) 0.1.a0 nocturnal flight call detector from the [BirdVox](https://wp.nyu.edu/birdvox/) project. Five different variants of the detector are available in Vesper, with nominal thresholds of 30, 40, 50, 60, and 70. All of the thresholds are adaptive, with a context duration of 30 minutes.

* Adds a workaround to the MPG Ranch tseep and thrush detectors concerning a problem that sometimes arises with multiple copies of the OpenMP runtime.

* Fixes a bug in the delete clips command on Windows. The bug caused the command to fail when attempting to delete clips for a night with approximately 1000 clips or more.

[](https://doi.org/10.5281/zenodo.2580460)

This release:

* Fixes two bugs concerning read-only archives.

* Adds evaluation scripts for the new MPG Ranch detectors.

[](https://doi.org/10.5281/zenodo.2563145)

This release:

* Adds two new detectors, MPG Ranch Tseep Detector 0.0 and MPG Ranch Thrush Detector 0.0. These detectors are preliminary versions of neural-network-based detectors trained on MPG Ranch data. In version 0.0, the neural networks are identical to those of the MPG Ranch NFC Coarse Classifier 3.0. In future versions the neural networks will be trained on data that better reflect the statistics of the audio segments that a detector encounters in a recording, as opposed to the segments that a coarse classifier encounters in clips.

* Improves the responsiveness of rapid navigation through clip album pages.

* Fixes a clip album bug that sometimes resulted in empty clip views after a settings change.

[](https://doi.org/10.5281/zenodo.2554614)

This release:

* Adds a new coarse classifier, MPG Ranch NFC Coarse Classifier 3.0, that annotates clips created by the Old Bird detectors as calls or noises. The new classifier performs considerably better than version 2.0.

* Updates MPG Ranch NFC Coarse Classifier 2.0 to version 2.1. The new version should label clips exactly as the old one did. It uses the version of Keras that is now built into TensorFlow rather than the separate version, however, eliminating dependence on the separate `keras` package.

* Increases the speed of clip album page loads by transferring clips from the server to the client in batches instead of one at a time.

* Increases the speed of the `Delete Clips` command.

* Updates Vesper to use the most recent version (2.1) of Django.

* Fixes broken archive data import drag and drop operations, issue 146.

* Adds a new time interval spectrogram clip view overlay.

* Adds a choice of marker type to the time-frequency point overlay.

[](https://doi.org/10.5281/zenodo.2434422)