# AudioCodec Benchmark (speaker verification)
Auther: Haibin Wu

## Resynthesis data (voxceleb evaluation data)
- The original data is here: https://huggingface.co/datasets/AudioDecBenchmark/Voxceleb1_test_original
- Resynthesis the data by different codecs: https://github.com/voidful/AudioDecBenchmark/blob/main/dataset_creator.py

## Run the speaker verificaiton for different codec resynthesised data
- git clone https://github.com/hbwu-ntu/ECAPA-TDNN.git
- download the resynthesised data to `your_path`
- Run the command

python trainECAPAModel.py --eval --initial_model exps/pretrain.model --eval_list exps/test.txt --eval_path ${your_path}