zhongshijun/beat-tracking-tcn

An implementation of Davies & Böck's beat-tracking temporal convolutional network

Beat Tracking TCN

An implementation of two adaptations to Davies & Böck's beat-tracking temporal convolutional network [1].

Usage

In order to use the beat tracker, this package and its dependencies must be installed with pip. It is recommended to do so in a virtualenv as follows:

python -m venv venv
source venv/bin/activate

The dependencies can then be installed like so*:

pip install -r requirements-env.txt && pip install -r requirements.txt

Once this is done, the beat tracker can be accessed like this:

from beat_tracking_tcn.beat_tracker import beatTracker

beats, downbeats = beatTracker("path_to_audio_file")

Alternatively, if you have the required dependencies librosa, madmom, mir-eval, numpy, and torch installed system-wide, you can perform a system wide install by running the following command from the root of this source repo:

pip install -e .

The beat tracker can then be invoked as above.

* It is important to use the full install command as listed, as madmom's setup.py itself depends on some packages (namely, cython and numpy. Splitting dependencies over two files like this prevents the install from falling down.

References

[1] M. E. P. Davies and S. Bock, ‘Temporal convolutional networks for musical audio beat tracking’, in 2019 27th European Signal Processing Conference (EUSIPCO), A Coruna, Spain, 2019, pp. 1–5, doi: 10.23919/EUSIPCO.2019.8902578.