francqz31 / vpidm

This is official repository for "variance-preserving-based interpolation diffusion models for speech enhanc"

（Variance Preserving Interpolation Diffusion Model for Speech Enhancement) VPIDM

variance-preserving-based interpolation diffusion models for speech enhancement, in which we apply the diffusion model to the speech enhancement (denoising) task. Diffusion Models for Speech Enhancement.

Preparation

Install requirements in requirements.txt via

pip install -r requirements.txt

Training

python train.py --base_dir <your vbd dataset dir>
                --gpus 4
                --no_wandb
                --sde vpsde
                --eta 1.5
                --beta-max 2
                --N 25
                --t_eps 4e-2
                --logdir <your log dir>

Enhancing

python enhancement.py --test_dir exp_dir
                      --corrector_step 0
                      --N 25
                      --enhanced_dir /outputs
                      --ckpt <your checkpoint best.ckpt>