This repo contains my solution to the Bandai Namco Data Science Challenge competition.
The objective of this competition was to remove the artificial noise from the given voice data. The voice data was converted into Mel Spectrogram.
raw[0-9]\{3}.npy
- Mel Spectrogram of raw voice data
- 100 files
noised_tgt_[0-9]{3}.npy
- Mel Spectrogram of noised voice data
- 30 files
I used a pix2pix model to achieve this competition.
- become-yukarin
- librosa
- for
convert_to_wav.py
- for
- numpy
- for
convert_to_wav.py
- for