Heavily inspired by this incredible YouTube channel. https://www.youtube.com/channel/UCZPFjMe1uRSirmSpznqvJfQ
Done:
- show spectrograms
- prepare the whole dataset into a stadard form, insipired by this video: https://www.youtube.com/watch?v=szyGiObZymo
- simple network that successfully compiles
- Fight overfitting
- Use LSTMs and ConvNet
Todo:
- Try melspectrograms instead of MFCC (or more features in MFCC, up to 40)
- Use more data, 2k may be not enough.
The dataset consists of 50 WAV files sampled at 16KHz for 50 different classes.
To each one of the classes, corresponds 40 audio sample of 5 seconds each. All of these audio files have been concatenated by class in order to have 50 wave files of 3 min. 20sec.
Download this to the same directory where this project is located.
run chmod +x ./preps.sh && ./preps.sh to start playing with the examples.