meldataset.py

import math
import os
import json
import random
import torch
# from torchvision.transforms.functional import resize
import torch.utils.data
import numpy as np
import librosa
from librosa.util import normalize
from scipy.io.wavfile import read
from librosa.filters import mel as librosa_mel_fn
# from speechbrain.lobes.models.FastSpeech2 import mel_spectogram

MAX_WAV_VALUE = 32768.0


def load_wav(full_path):
    sampling_rate, data = read(full_path)
    return data, sampling_rate


def dynamic_range_compression(x, C=1, clip_val=1e-5):
    return np.log(np.clip(x, a_min=clip_val, a_max=None) * C)


def dynamic_range_decompression(x, C=1):
    return np.exp(x) / C


def dynamic_range_compression_torch(x, C=1, clip_val=1e-5):
    return torch.log(torch.clamp(x, min=clip_val) * C)


def dynamic_range_decompression_torch(x, C=1):
    return torch.exp(x) / C


def spectral_normalize_torch(magnitudes):
    output = dynamic_range_compression_torch(magnitudes)
    return output


def spectral_de_normalize_torch(magnitudes):
    output = dynamic_range_decompression_torch(magnitudes)
    return output


mel_basis = {}
hann_window = {}


def mel_spectrogram(y, n_fft, num_mels, sampling_rate, hop_size, win_size, fmin, fmax, center=False):
    # if torch.min(y) < -1.:
    #     print('min value is ', torch.min(y))
    # if torch.max(y) > 1.:
    #     print('max value is ', torch.max(y))

    global mel_basis, hann_window
    if fmax not in mel_basis:
        mel = librosa_mel_fn(sr=sampling_rate, n_fft=n_fft, n_mels=num_mels, fmin=fmin, fmax=fmax)
        mel_basis[str(fmax)+'_'+str(y.device)] = torch.from_numpy(mel).float().to(y.device)
        hann_window[str(y.device)] = torch.hann_window(win_size).to(y.device)

    y = torch.nn.functional.pad(y.unsqueeze(1), (int((n_fft-hop_size)/2), int((n_fft-hop_size)/2)), mode='reflect')
    y = y.squeeze(1)

    # complex tensor as default, then use view_as_real for future pytorch compatibility
    spec = torch.stft(y, n_fft, hop_length=hop_size, win_length=win_size, window=hann_window[str(y.device)],
                      center=center, pad_mode='reflect', normalized=False, onesided=True, return_complex=True)
    spec = torch.view_as_real(spec)
    spec = torch.sqrt(spec.pow(2).sum(-1)+(1e-9))

    spec = torch.matmul(mel_basis[str(fmax)+'_'+str(y.device)], spec)
    spec = spectral_normalize_torch(spec)

    return spec


def get_dataset_filelist(a):
    training_files =[]
    validation_files =[]
    total_files = 0

    audio_dir = "dataset/audio"

    with open("filelists/train.txt") as f:
        training_files = f.readlines()
    for i, line in enumerate(training_files):
        spk, basename = line.strip().split('|')
        training_files[i] = f"{audio_dir}/{spk}/{basename}.wav"

    with open("filelists/val.txt") as f:
        validation_files = f.readlines()
    for i, line in enumerate(validation_files):
        spk, basename = line.strip().split('|')
        validation_files[i] = f"{audio_dir}/{spk}/{basename}.wav"
    
    random.seed(1234)
    random.shuffle(training_files)
    random.shuffle(validation_files)

    return training_files, validation_files


class MelDataset(torch.utils.data.Dataset):
    def __init__(self, training_files, segment_size, n_fft, num_mels,
                 hop_size, win_size, sampling_rate,  fmin, fmax, shuffle=True, n_cache_reuse=1,
                 device=None, fmax_loss=None, use_aug=False):
        self.audio_files = training_files
        random.seed(1234)
        if shuffle:
            random.shuffle(self.audio_files)
        self.segment_size = segment_size
        self.sampling_rate = sampling_rate
        self.n_fft = n_fft
        self.num_mels = num_mels
        self.hop_size = hop_size
        self.win_size = win_size
        self.fmin = fmin
        self.fmax = fmax
        self.fmax_loss = fmax_loss
        self.cached_wav = None
        self.n_cache_reuse = n_cache_reuse
        self._cache_ref_count = 0
        self.device = device
        self.use_aug = use_aug

        with open("filelists/spk2id.json") as f:
            self.spk2id = json.load(f)

    def __getitem__(self, index):
        filename = self.audio_files[index]
        if self._cache_ref_count == 0:
            audio, sampling_rate = load_wav(filename)
            audio = audio / MAX_WAV_VALUE
            audio = normalize(audio) * 0.95
            self.cached_wav = audio
            if sampling_rate != self.sampling_rate:
                raise ValueError("{} SR doesn't match target {} SR".format(
                    sampling_rate, self.sampling_rate))
            self._cache_ref_count = self.n_cache_reuse
        else:
            audio = self.cached_wav
            self._cache_ref_count -= 1

        audio = torch.FloatTensor(audio)
        audio = audio.unsqueeze(0)

        if audio.size(1) >= self.segment_size:
            max_audio_start = audio.size(1) - self.segment_size
            audio_start = random.randint(0, max_audio_start)
            audio = audio[:, audio_start:audio_start+self.segment_size]
        else:
            audio = torch.nn.functional.pad(audio, (0, self.segment_size - audio.size(1)), 'constant')

        mel = mel_spectrogram(audio, self.n_fft, self.num_mels,
                                self.sampling_rate, self.hop_size, self.win_size, self.fmin, self.fmax,
                                center=False)

        mel_loss = mel_spectrogram(audio, self.n_fft, self.num_mels,
                                   self.sampling_rate, self.hop_size, self.win_size, self.fmin, self.fmax_loss,
                                   center=False)
        
        spk_path = filename.replace("audio", "spk").replace(".wav", ".npy")
        spk_emb = torch.from_numpy(np.load(spk_path)) # (256)
        spk = filename.split("/")[-2]
        spk_id = self.spk2id[spk]
        spk_id = torch.LongTensor([spk_id])

        if not self.use_aug:
            return (mel.squeeze(), audio.squeeze(0), filename, mel_loss.squeeze(), spk_emb, spk_id)
        
        mel_aug, _ = mel_spectogram(
            audio=audio.squeeze(),
            sample_rate=16000,
            hop_length=256,
            win_length=1024,
            n_mels=80,
            n_fft=1024,
            f_min=0.0,
            f_max=8000.0,
            power=1,
            normalized=False,
            min_max_energy_norm=True,
            norm="slaney",
            mel_scale="slaney",
            compression=True
        )
        mel_aug = self.resize_mel(mel_aug.unsqueeze(0)).squeeze(0)

        return (mel_aug.squeeze(), mel.squeeze(), audio.squeeze(0), filename, mel_loss.squeeze(), spk_emb, spk_id)

    def __len__(self):
        return len(self.audio_files)
    
    def resize_mel(self, mel):
        ratio = 0.85 + 0.3 * torch.rand(1) # 0.85 ~ 1.15
        height = int(mel.size(-2) * ratio)
        width = mel.size(-1)
        
        mel_r = resize(mel, (height, width), antialias=True)
        
        if height >= mel.size(-2):
            mel_r = mel_r[:, :mel.size(-2), :]
        else:
            pad = mel_r[:, -1:, :].repeat(1, mel.size(-2) - height, 1) 
            pad += torch.randn_like(pad) / 1e3
            mel_r = torch.cat((mel_r, pad), 1)
        
        return mel_r