Variational autoencoder for audio
Budget: €8 – €30 EUR
Take a Variational Autoencoder, and replace the convolutional layers by RNN layers (Version 0) and by GRU layers (Version 1).
Have 2 layers in the encoder and 2 layers in the decoder part. Use a donw sampling factor of 8 after the encoder.
Train it on music and speech files , and use a separate set for testing.
Then evaluate the audio quality after decoding
Have 2 layers in the encoder and 2 layers in the decoder part. Use a donw sampling factor of 8 after the encoder.
Train it on music and speech files , and use a separate set for testing.
Then evaluate the audio quality after decoding