Neural networks applied in recognizing guitar chords using python, AutoML.NET with C# and .NET Core
The demo application is written in C# with .NETCore. As of July 9, 2020, the only version available is for windows 10 64 bits. Versions for Linux are expected to come as a console application.
Installing the application.
The demo app uses AutoML .NET as the default prediction engine and ONNX runtime as the legacy prediction engine to run the exported model created on python and keras and it needs Visual C++ to be installed on the machine that is going to run the app.
- (Optional, only if you are interested on using ONNX runtime) Install Visual C++ from the Microsoft web site
- Download the application from our Releases
- Extract the folder and run
- Load any .WAV or .MP3 file and it will split your file in different chords
- Play, Pause and Stop audio controls
- Seek to any chord in particular and resume the reproduction from there.
- Ability to change the length of the window for analyzing the chords
- Ability to correct the model prediction
- Retrain the model based on your corrections
NOTE this was tested using the following setup:
os: Ubuntu 18.04 python --version Python 3.6.9
This program is a university project on the introductory course to artificial intelligence.
You'll need python 3, pip, and virtualenv(optional but recommended) to run the program
- Clone the repository
- Go inside the project folder
- Create a virtualenv
virtualenv -p python3 my_env
- Activate your environment
- Install the dependencies
pip install -r requirements.txt
At this point you have the environment ready to use several entry points.
python predict.py path/to/song.wavwill give you a prediction of the chord present in that sound file, by default it goes to
songs/d.wavwhich is the D chord. It will show a window similar to this one.
python split.py songs/guitar/about_a_girl.wavis an example of the
split.pyentry point that takes a longer song, splits it and runs the prediction for each song piece. The results are saved in a filed called
- Example, if you ran
python split.py songs/guitar/about_a_girl.wav, then
spliter_result.txtwould have the following content:
em g em g em em em g em em g g em g g g em g g g em em g em em g g em em g gwhich are the chords that it was able to identify on the song.
- Example, if you ran
In the file
paper.pdfyou'll find the final report of this university project with some references added. Currently it is only available in Spanish, you could go and try to use a translator, hope it helps