TOLD/SOND: add utt2num_frame script

志浩

2023-08-02 a8701ad5df6255502c431fb749b4661ff4853e57

TOLD/SOND: add utt2num_frame script

 egs/callhome/diarization/sond/finetune.sh

@@ -180,9 +180,11 @@
    git lfs install
    git clone https://www.modelscope.cn/damo/speech_xvector_sv-en-us-callhome-8k-spk6135-pytorch.git
    mv speech_xvector_sv-en-us-callhome-8k-spk6135-pytorch ${expdir}/
    echo "Done."
  fi

  for dset in callhome1/nonoverlap_0s callhome2/nonoverlap_0s; do
    echo "Start to extract speaker embeddings for ${dset}"
    key_file=${datadir}/${dset}/wav.scp
    num_scp_file="$(<${key_file} wc -l)"
    _nj=$([ $inference_nj -le $num_scp_file ] && echo "$inference_nj" || echo "$num_scp_file")
@@ -207,6 +209,9 @@
        --sv_model_file ${sv_exp_dir}/sv.pth \
        --output_dir "${_logdir}"/output.JOB
    cat ${_logdir}/output.*/xvector.scp | sort > ${datadir}/${dset}/utt2xvec

    python script/calc_num_frames.py ${key_file} ${datadir}/${dset}/utt2num_frames
    echo "Done."
  done

fi
@@ -219,7 +224,7 @@
    python -Wignore script/calc_real_meeting_frame_labels.py \
          ${datadir}/${dset} ${dumpdir}/${dset}/labels \
          --n_spk 8 --frame_shift 0.01 --nj 16 --sr 8000
    find `pwd`/${dumpdir}/${dset}/labels -iname "*.lbl.mat" | awk -F'[/.]' '{print $(NF-2),$0}' | sort > ${datadir}/${dset}/labels.scp
    find `pwd`/${dumpdir}/${dset}/labels/ -iname "*.lbl.mat" | awk -F'[/.]' '{print $(NF-2),$0}' | sort > ${datadir}/${dset}/labels.scp
  done

fi

 egs/callhome/diarization/sond/script/calc_num_frames.py

New file
@@ -0,0 +1,21 @@
import os
import sys
import soundfile as sf
from funasr.utils.misc import load_scp_as_list


if __name__ == '__main__':
    wav_scp = sys.argv[1]
    out_file = sys.argv[2]
    frame_shift = 0.01

    os.makedirs(os.path.dirname(out_file), exist_ok=True)

    out_file = open(out_file, "wt")
    for uttid, wav_path in load_scp_as_list(wav_scp):
        wav, sr = sf.read(wav_path)
        num_frame = wav.shape[0] // int(sr * frame_shift)
        out_file.write(f"{uttid} {num_frame}\n")
        out_file.flush()

    out_file.close()

 egs/callhome/diarization/sond/script/calc_real_meeting_frame_labels.py

@@ -1,6 +1,6 @@
import numpy as np
from opennmt.utils.job_runner import MultiProcessRunnerV3
from opennmt.utils.misc import load_scp_as_list, load_scp_as_dict
from funasr.utils.job_runner import MultiProcessRunnerV3
from funasr.utils.misc import load_scp_as_list, load_scp_as_dict
import os
import librosa
import scipy.io as sio
@@ -90,7 +90,7 @@
    for mid, wav_path, rttms in task_list:
        meeting_labels, spk_list = build_labels(wav_path, rttms, args.n_spk, args.remove_sil,
                                                args.sr, args.frame_shift)
        save_path = os.path.join(args.out_dir, "{}.lbl".format(mid))
        save_path = os.path.join(args.out_dir, "{}.lbl.mat".format(mid))
        sio.savemat(save_path, {"labels": meeting_labels.astype(bool), "spk_list": spk_list})
        # print mid
    return None