Step 0: Preparation

① Clone the repository and enter the project directory

git clone https://github.com/REAL-TSE/REAL-T.git
cd REAL-T

② Create a Conda environment and install dependencies

conda create -n REAL-T python=3.9
conda activate REAL-T
pip install -r requirements.txt

# install wesep
git submodule init
git submodule update

③ Set up environment variables

Please replace $PWD below with the absolute path to this project.

export PATH=$PWD/FireRedASR/fireredasr/:$PWD/FireRedASR/fireredasr/utils/:$PATH
export PYTHONPATH=$PWD/FireRedASR/:$PYTHONPATH
export PYTHONPATH=$PWD/wesep:$PYTHONPATH

④ Automatically prepare dataset and checkpoints

bash -i ./pre.sh

Step 1: Run TSE Inference

Run inference script (You can adapt its input/output structure to suit your own TSE model)

bash -i ./run_tse.sh

Step 2: ASR-based Evaluation

Run both transcription and evaluation

bash -i ./transcribe_and_evaluation.sh 1 2

Performance

The table below compares the performance of several recently proposed TSE models on the simulated Libri2Mix and PRIMARY test sets.