Poly Pitch Net

Install python packages

setup python venv

python3.10 -m venv venv
source venv/bin/activate

pip install amt-tools
# install guitar-transcription-continuous
pip install git+https://github.com/cwitkowitz/guitar-transcription-continuous.git@d481054f54184374c04b1cc27a487dc35c87f353
# install guitar-transcription-inhibition
pip install git+https://github.com/cwitkowitz/guitar-transcription-with-inhibition.git@e611c1dc9b7340d35c9a697d1658b3b2afb3978a

Experiments

poly_pitch_net_experiments directory contains scripts with the implementation of training, inference and valuation of the Poly Pitch Net models with the use of the GuitarSet dataset. When running the experiment.py script for the first time, make sure that the ../generated/ directory is empty - this is the cache directory for all the validation and training sets. If any errors related to missing dictionary keys occur, make sure that the reset_data flag in the GuitarSet init calls is set to True.

run the training script

python3 poly_pitch_net_experiments/experiment_fretnet_cnn.py

run the training script with nohup

nohup python3 poly_pitch_net_experiments/experiment_fretnet_cnn.py > output.txt 2>&1 &

Tests

Pytest is the framework of our choice.

pip install pytest

To run the tests, first, install the poly_pitch_net package as a developement package, run below in the root of the poly-pitch-net git project directory.

pip install -e .

Finally, run the tests

pytest -v