philippemos's picture
change num_worker in nemo configs to enforce cpu usage
6bfc1e0
raw
history blame contribute delete
No virus
6.28 kB
"""
General streamlit diarization application
"""
import os
import shutil
from io import BytesIO
from typing import Dict, Union
from pathlib import Path
import librosa
import librosa.display
import matplotlib.figure
import numpy as np
import streamlit as st
import streamlit.uploaded_file_manager
from PIL import Image
from pydub import AudioSegment
from matplotlib import pyplot as plt
import configs
from utils import audio_utils, text_utils, general_utils, streamlit_utils
from diarizers import pyannote_diarizer, nemo_diarizer
plt.rcParams["figure.figsize"] = (10, 5)
def plot_audio_diarization(diarization_figure: Union[plt.gcf, np.array], diarization_name: str,
audio_data: np.array,
sampling_frequency: int):
"""
Function that plots the audio along with the different applied diarization techniques
Args:
diarization_figure (plt.gcf): the diarization figure to plot
diarization_name (str): the name of the diarization technique
audio_data (np.array): the audio numpy array
sampling_frequency (int): the audio sampling frequency
"""
col1, col2 = st.columns([3, 5])
with col1:
st.markdown(
f"<h4 style='text-align: center; color: black;'>Original</h5>",
unsafe_allow_html=True,
)
st.markdown("<br></br>", unsafe_allow_html=True)
st.audio(audio_utils.create_st_audio_virtualfile(audio_data, sampling_frequency))
with col2:
st.markdown(
f"<h4 style='text-align: center; color: black;'>{diarization_name}</h5>",
unsafe_allow_html=True,
)
if type(diarization_figure) == matplotlib.figure.Figure:
buf = BytesIO()
diarization_figure.savefig(buf, format="png")
st.image(buf)
else:
st.image(diarization_figure)
st.markdown("---")
def execute_diarization(file_uploader: st.uploaded_file_manager.UploadedFile, selected_option: any,
sample_option_dict: Dict[str, str],
diarization_checkbox_dict: Dict[str, bool],
session_id: str):
"""
Function that exectutes the diarization based on the specified files and pipelines
Args:
file_uploader (st.uploaded_file_manager.UploadedFile): the uploaded streamlit audio file
selected_option (any): the selected option of samples
Dict[str, str]: a dictionary where the name is the file name (without extension to be listed
as an option for the user) and the value is the original file name
diarization_checkbox_dict (Dict[str, bool]): dictionary where the key is the Diarization
technique name and the value is a boolean indicating whether to apply that technique
session_id (str): unique id of the user session
"""
user_folder = os.path.join(configs.UPLOADED_AUDIO_SAMPLES_DIR, session_id)
Path(user_folder).mkdir(parents=True, exist_ok=True)
if file_uploader is not None:
file_name = file_uploader.name
file_path = os.path.join(user_folder, file_name)
audio = AudioSegment.from_wav(file_uploader).set_channels(1)
# slice first 30 seconds (slicing is done by ms)
audio = audio[0:1000 * 30]
audio.export(file_path, format='wav')
else:
file_name = sample_option_dict[selected_option]
file_path = os.path.join(configs.AUDIO_SAMPLES_DIR, file_name)
audio_data, sampling_frequency = librosa.load(file_path)
nb_pipelines_to_run = sum(pipeline_bool for pipeline_bool in diarization_checkbox_dict.values())
pipeline_count = 0
for diarization_idx, (diarization_name, diarization_bool) in \
enumerate(diarization_checkbox_dict.items()):
if diarization_bool:
pipeline_count += 1
if diarization_name == 'pyannote':
diarizer = pyannote_diarizer.PyannoteDiarizer(file_path)
elif diarization_name == 'NeMo':
diarizer = nemo_diarizer.NemoDiarizer(file_path, user_folder)
else:
raise NotImplementedError('Framework not recognized')
if file_uploader is not None:
with st.spinner(
f"Executing {pipeline_count}/{nb_pipelines_to_run} diarization pipelines "
f"({diarization_name}). This might take 1-2 minutes..."):
diarizer_figure = diarizer.get_diarization_figure()
else:
diarizer_figure = Image.open(f"{configs.PRECOMPUTED_DIARIZATION_FIGURE}/"
f"{file_name.rsplit('.')[0]}_{diarization_name}.png")
plot_audio_diarization(diarizer_figure, diarization_name, audio_data,
sampling_frequency)
shutil.rmtree(user_folder)
st.set_page_config(
page_title="πŸ“œ Audio diarization visualization πŸ“œ",
page_icon="",
layout="wide",
initial_sidebar_state="auto",
menu_items={
'Get help': None,
'Report a bug': None,
'About': None,
}
)
text_utils.intro_container()
# 2.1) Diarization method
text_utils.demo_container()
st.markdown("Choose the Diarization method here:")
diarization_checkbox_dict = {}
for diarization_method in configs.DIARIZATION_METHODS:
diarization_checkbox_dict[diarization_method] = st.checkbox(
diarization_method)
# 2.2) Diarization upload/sample select
st.markdown("(Optional) Upload an audio file here:")
file_uploader = st.file_uploader(
label="", type=[".wav", ".wave"]
)
sample_option_dict = general_utils.get_dict_of_audio_samples(configs.AUDIO_SAMPLES_DIR)
st.markdown("Or select a sample file here:")
selected_option = st.selectbox(
label="", options=list(sample_option_dict.keys())
)
st.markdown("---")
## 2.3) Apply specified diarization pipeline
if st.button("Apply"):
session_id = streamlit_utils.get_session()
execute_diarization(
file_uploader=file_uploader,
selected_option=selected_option,
sample_option_dict=sample_option_dict,
diarization_checkbox_dict=diarization_checkbox_dict,
session_id=session_id
)
text_utils.conlusion_container()