Implementation of paper "End-to-end lyrics alignment for polyphonic music using an audio-to-character recognition model" (link) based on the pytorch implementation of Wave-U-Net.
The input and output size are reduced to stablize training on the DALI dataset.
To train the model, pull the DALI wrapper somewhere and link it from the root of this repository.
ln -s path/to/dali_wrapper/ DALI
Then run the following command. Set --cuda flag if you have GPUs.
python train.py --dataset_dir path/to/DALI_v2.0/