DeepSuppressor is a deep learning-based speech denoiser which can significantly improve the quality of speech signals. The denoiser is trained on a large dataset of clean and noisy speech signals, and it can be used to denoise speech signals in real time or offline.
To get started with this project, follow the steps below:
Clone the repository
-
Clone the repository to your local machine using the following command:
git clone https://github.com/muhd-umer/deep-suppressor.git
Create a new virtual environment
-
It is recommended to create a new virtual environment so that updates/downgrades of packages do not break other projects. To create a new virtual environment, run the following command:
conda env create -f environment.yml
-
Alternatively, you can use
mamba
(faster than conda) package manager to create a new virtual environment:wget -O miniforge.sh \ "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh" bash miniforge.sh -b -p "${HOME}/conda" source "${HOME}/conda/etc/profile.d/conda.sh" # For mamba support also run the following command source "${HOME}/conda/etc/profile.d/mamba.sh" conda activate mamba env create -f environment.yml
Install the dependencies
-
Activate the newly created environment:
conda activate deep-suppressor
-
Install PyTorch (Stable 2.0.1):
pip3 install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118