app.py

import numpy as np

import streamlit as st
import librosa
import soundfile as sf
import librosa.display
from config import CONFIG
import torch
from dataset import MaskGenerator
import onnxruntime, onnx
import matplotlib.pyplot as plt
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
from torchmetrics.audio.stoi import ShortTimeObjectiveIntelligibility as STOI
from torchmetrics.audio.pesq import PerceptualEvaluationSpeechQuality as PESQ
import pandas as pd
import torchaudio

@st.cache
def load_model():
    path = 'lightning_logs/version_0/checkpoints/frn.onnx'
    onnx_model = onnx.load(path)
    options = onnxruntime.SessionOptions()
    options.intra_op_num_threads = 2
    options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL
    session = onnxruntime.InferenceSession(path, options)
    input_names = [x.name for x in session.get_inputs()]
    output_names = [x.name for x in session.get_outputs()]
    return session, onnx_model, input_names, output_names

def inference(re_im, session, onnx_model, input_names, output_names):
    inputs = {input_names[i]: np.zeros([d.dim_value for d in _input.type.tensor_type.shape.dim],
                                       dtype=np.float32)
              for i, _input in enumerate(onnx_model.graph.input)
              }

    output_audio = []
    for t in range(re_im.shape[0]):
        inputs[input_names[0]] = re_im[t]
        out, prev_mag, predictor_state, mlp_state = session.run(output_names, inputs)
        inputs[input_names[1]] = prev_mag
        inputs[input_names[2]] = predictor_state
        inputs[input_names[3]] = mlp_state
        output_audio.append(out)

    output_audio = torch.tensor(np.concatenate(output_audio, 0))
    output_audio = output_audio.permute(1, 0, 2).contiguous()
    output_audio = torch.view_as_complex(output_audio)
    output_audio = torch.istft(output_audio, window, stride, window=hann)
    return output_audio.numpy()

def visualize(hr, lr, recon, sr):
    sr = sr
    window_size = 1024
    window = np.hanning(window_size)

    stft_hr = librosa.core.spectrum.stft(hr, n_fft=window_size, hop_length=512, window=window)
    stft_hr = 2 * np.abs(stft_hr) / np.sum(window)

    stft_lr = librosa.core.spectrum.stft(lr, n_fft=window_size, hop_length=512, window=window)
    stft_lr = 2 * np.abs(stft_lr) / np.sum(window)

    stft_recon = librosa.core.spectrum.stft(recon, n_fft=window_size, hop_length=512, window=window)
    stft_recon = 2 * np.abs(stft_recon) / np.sum(window)

    fig, (ax1, ax2, ax3) = plt.subplots(3, 1, sharey=True, sharex=True, figsize=(16, 12))
    ax1.title.set_text('Оригинальный сигнал')
    ax2.title.set_text('Сигнал с потерями')
    ax3.title.set_text('Улучшенный сигнал')

    canvas = FigureCanvas(fig)
    p = librosa.display.specshow(librosa.amplitude_to_db(stft_hr), ax=ax1, y_axis='log', x_axis='time', sr=sr)
    p = librosa.display.specshow(librosa.amplitude_to_db(stft_lr), ax=ax2, y_axis='log', x_axis='time', sr=sr)
    p = librosa.display.specshow(librosa.amplitude_to_db(stft_recon), ax=ax3, y_axis='log', x_axis='time', sr=sr)
    return fig

packet_size = CONFIG.DATA.EVAL.packet_size
window = CONFIG.DATA.window_size
stride = CONFIG.DATA.stride

title = 'Сокрытие потерь пакетов'
st.set_page_config(page_title=title, page_icon=":sound:")
st.title(title)

st.subheader('1. Загрузка аудио')
uploaded_file = st.file_uploader("Загрузите аудио формата (.wav) 48 КГц")

is_file_uploaded = uploaded_file is not None
if not is_file_uploaded:
    uploaded_file = 'sample.wav'

target, sr = librosa.load(uploaded_file)
target = target[:packet_size * (len(target) // packet_size)]

st.text('Ваше аудио')
st.audio(uploaded_file)

st.subheader('2. Выберите желаемый процент потерь')
slider = [st.slider("Ожидаемый процент потерь для генератора потерь цепи Маркова", 0, 100, step=1)]
loss_percent = float(slider[0])/100
mask_gen = MaskGenerator(is_train=False, probs=[(1 - loss_percent, loss_percent)])
lossy_input = target.copy().reshape(-1, packet_size)
mask = mask_gen.gen_mask(len(lossy_input), seed=0)[:, np.newaxis]
lossy_input *= mask
lossy_input = lossy_input.reshape(-1)
hann = torch.sqrt(torch.hann_window(window))
lossy_input_tensor = torch.tensor(lossy_input)
re_im = torch.stft(lossy_input_tensor, window, stride, window=hann, return_complex=False).permute(1, 0, 2).unsqueeze(
    1).numpy().astype(np.float32)
session, onnx_model, input_names, output_names = load_model()

if st.button('Сгенерировать потери'):
    with st.spinner('Ожидайте...'):
        output = inference(re_im, session, onnx_model, input_names, output_names)

        st.subheader('3. Визуализация')
        fig = visualize(target, lossy_input, output, sr)
        st.pyplot(fig)
    st.success('Сделано!')
    sf.write('target.wav', target, sr)
    sf.write('lossy.wav', lossy_input, sr)
    sf.write('enhanced.wav', output, sr)
    st.text('Оригинальное аудио')
    st.audio('target.wav')
    st.text('Аудио с потерями')
    st.audio('lossy.wav')
    st.text('Улучшенное аудио')
    st.audio('enhanced.wav')
    data_clean, samplerate = torchaudio.load('target.wav')

    data_lossy, samplerate = torchaudio.load('lossy.wav')
    data_enhanced, samplerate = torchaudio.load('enhanced.wav')
    min_len = min(data_clean.shape[1], data_lossy.shape[1], data_enhanced.shape[1])
    data_clean = data_clean[:, :min_len]
    data_lossy = data_lossy[:, :min_len]
    data_enhanced = data_enhanced[:, :min_len]


    stoi = STOI(samplerate)
    stoi_orig = round(float(stoi(data_clean, data_clean)),3)
    stoi_lossy = round(float(stoi(data_clean, data_lossy)),5)
    stoi_enhanced = round(float(stoi(data_clean, data_enhanced)),5)
    stoi_mass=[stoi_orig, stoi_lossy, stoi_enhanced]


    #pesq = PESQ(16000, 'nb')
    #data_clean = data_clean.cpu().numpy()
    #data_lossy = data_lossy.detach().cpu().numpy()
    #data_enhanced = data_enhanced.cpu().numpy()

    #if samplerate != 16000:
    #    data_lossy = librosa.resample(data_lossy, orig_sr=48000, target_sr=16000)
    #    data_clean = librosa.resample(data_clean, orig_sr=48000, target_sr=16000)
    #    data_enhanced = librosa.resample(data_enhanced, orig_sr=48000, target_sr=16000)

    #    pesq_orig = np.array(pesq(torch.tensor(data_clean), torch.tensor(data_clean)))
    #    pesq_lossy = np.array(pesq(torch.tensor(data_lossy), torch.tensor(data_clean)))
    #    pesq_enhanced = np.array(pesq(torch.tensor(data_enhanced), torch.tensor(data_clean)))

    #psq_mas=[pesq_orig, pesq_lossy, pesq_enhanced]



    df = pd.DataFrame(columns=['Audio', 'PESQ', 'STOI', 'PLCMOS', 'LSD'])

    df['Audio'] = ['Clean', 'Lossy', 'Enhanced']

    #df['PESQ'] = psq_mas

    df['STOI'] = stoi_mass

    st.table(df)