Training for wav to midi transcriper

[mito0o852]

I’m starting an experiment to train the model to converts music audio to MIDI. The approach involves freezing the encoder and continuing training on the decoder. Any suggestions or feedback before I begin?

[jishengpeng]

I’m starting an experiment to train the model to converts music audio to MIDI. The approach involves freezing the encoder and continuing training on the decoder. Any suggestions or feedback before I begin?

The sampling rate of WavTokenizer for both audio and music is relatively low, at only 24k. I am unsure whether this might limit the generation of high-quality MIDI.