heimoshuiyu/ultimatevocalremovergui
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add files via upload

Browse Source
This commit is contained in:
Anjok07
2022-04-12 03:18:35 -05:00
committed by GitHub
parent b702ef99ad
commit 1edb2bec17
1 changed files with 104 additions and 94 deletions
Download Patch File Download Diff File Expand all files Collapse all files

198
inference_v5.py
View File

@@ -1,7 +1,16 @@
from functools import total_ordering
import pprint import pprint
import argparse import argparse
import os import os
import importlib import importlib
from statistics import mode
import sys
import subprocess
import contextlib
from subprocess import run
from tkinter.ttk import Progressbar
from typing import _SpecialForm, overload
from unittest.mock import _SpecState
import cv2 import cv2
import librosa import librosa
@@ -20,9 +29,7 @@ from collections import defaultdict
import tkinter as tk import tkinter as tk
import traceback # Error Message Recent Calls import traceback # Error Message Recent Calls
import time # Timer import time # Timer
import random
class VocalRemover(object): class VocalRemover(object):
@@ -44,7 +51,10 @@ class VocalRemover(object):
global args global args
global model_params_d global model_params_d
#progressb = tqdm
p = argparse.ArgumentParser() p = argparse.ArgumentParser()
p.add_argument('--paramone', type=str, default='lib_v5/modelparams/4band_44100.json') p.add_argument('--paramone', type=str, default='lib_v5/modelparams/4band_44100.json')
p.add_argument('--paramtwo', type=str, default='lib_v5/modelparams/4band_v2.json') p.add_argument('--paramtwo', type=str, default='lib_v5/modelparams/4band_v2.json')
@@ -96,85 +106,6 @@ class VocalRemover(object):
self.text_widget.write('Done!\n') self.text_widget.write('Done!\n')
def _execute(self, X_mag_pad, roi_size, n_window, device, model, aggressiveness):
model.eval()
with torch.no_grad():
preds = []
for i in tqdm(range(n_window)):
start = i * roi_size
X_mag_window = X_mag_pad[None, :, :, start:start + self.data['window_size']]
X_mag_window = torch.from_numpy(X_mag_window).to(device)
pred = model.predict(X_mag_window, aggressiveness)
pred = pred.detach().cpu().numpy()
preds.append(pred[0])
pred = np.concatenate(preds, axis=2)
return pred
def preprocess(self, X_spec):
X_mag = np.abs(X_spec)
X_phase = np.angle(X_spec)
return X_mag, X_phase
def inference(self, X_spec, device, model, aggressiveness):
X_mag, X_phase = self.preprocess(X_spec)
coef = X_mag.max()
X_mag_pre = X_mag / coef
n_frame = X_mag_pre.shape[2]
pad_l, pad_r, roi_size = dataset.make_padding(n_frame,
self.data['window_size'], model.offset)
n_window = int(np.ceil(n_frame / roi_size))
X_mag_pad = np.pad(
X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
pred = self._execute(X_mag_pad, roi_size, n_window,
device, model, aggressiveness)
pred = pred[:, :, :n_frame]
return pred * coef, X_mag, np.exp(1.j * X_phase)
def inference_tta(self, X_spec, device, model, aggressiveness):
X_mag, X_phase = self.preprocess(X_spec)
coef = X_mag.max()
X_mag_pre = X_mag / coef
n_frame = X_mag_pre.shape[2]
pad_l, pad_r, roi_size = dataset.make_padding(n_frame,
self.data['window_size'], model.offset)
n_window = int(np.ceil(n_frame / roi_size))
X_mag_pad = np.pad(
X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
pred = self._execute(X_mag_pad, roi_size, n_window,
device, model, aggressiveness)
pred = pred[:, :, :n_frame]
pad_l += roi_size // 2
pad_r += roi_size // 2
n_window += 1
X_mag_pad = np.pad(
X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
pred_tta = self._execute(X_mag_pad, roi_size, n_window,
device, model, aggressiveness)
pred_tta = pred_tta[:, :, roi_size // 2:]
pred_tta = pred_tta[:, :, :n_frame]
return (pred + pred_tta) * 0.5 * coef, X_mag, np.exp(1.j * X_phase)
data = { data = {
# Paths # Paths
'input_paths': None, 'input_paths': None,
@@ -230,9 +161,9 @@ def determineModelFolderName():
return modelFolderName return modelFolderName
def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress_var: tk.Variable, def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress_var: tk.Variable,
**kwargs: dict): **kwargs: dict):
def save_files(wav_instrument, wav_vocals): def save_files(wav_instrument, wav_vocals):
"""Save output music files""" """Save output music files"""
vocal_name = '(Vocals)' vocal_name = '(Vocals)'
@@ -350,23 +281,102 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
step=0.1) step=0.1)
text_widget.write(base_text + 'Stft of wave source...\n') text_widget.write(base_text + 'Stft of wave source...\n')
text_widget.write(base_text + 'Done!\n')
text_widget.write(base_text + "Please Wait..\n")
X_spec_m = spec_utils.combine_spectrograms(X_spec_s, mp) X_spec_m = spec_utils.combine_spectrograms(X_spec_s, mp)
del X_wave, X_spec_s del X_wave, X_spec_s
if data['tta']: def inference(X_spec, device, model, aggressiveness):
pred, X_mag, X_phase = vocal_remover.inference_tta(X_spec_m,
device, def _execute(X_mag_pad, roi_size, n_window, device, model, aggressiveness):
model, {'value': args.aggressiveness,'split_bin': mp.param['band'][1]['crop_stop']}) model.eval()
else:
pred, X_mag, X_phase = vocal_remover.inference(X_spec_m, with torch.no_grad():
device, preds = []
model, {'value': args.aggressiveness,'split_bin': mp.param['band'][1]['crop_stop']})
iterations = [n_window]
total_iterations = sum(iterations)
text_widget.write(base_text + "Length: "f"{total_iterations} Slices\n")
for i in tqdm(range(n_window)):
update_progress(**progress_kwargs,
step=(0.1 + (0.8/n_window * i)))
start = i * roi_size
X_mag_window = X_mag_pad[None, :, :, start:start + data['window_size']]
X_mag_window = torch.from_numpy(X_mag_window).to(device)
pred = model.predict(X_mag_window, aggressiveness)
pred = pred.detach().cpu().numpy()
preds.append(pred[0])
pred = np.concatenate(preds, axis=2)
return pred
def preprocess(X_spec):
X_mag = np.abs(X_spec)
X_phase = np.angle(X_spec)
return X_mag, X_phase
X_mag, X_phase = preprocess(X_spec)
coef = X_mag.max()
X_mag_pre = X_mag / coef
n_frame = X_mag_pre.shape[2]
pad_l, pad_r, roi_size = dataset.make_padding(n_frame,
data['window_size'], model.offset)
n_window = int(np.ceil(n_frame / roi_size))
X_mag_pad = np.pad(
X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
pred = _execute(X_mag_pad, roi_size, n_window,
device, model, aggressiveness)
pred = pred[:, :, :n_frame]
if data['tta']:
pad_l += roi_size // 2
pad_r += roi_size // 2
n_window += 1
X_mag_pad = np.pad(
X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
pred_tta = _execute(X_mag_pad, roi_size, n_window,
device, model, aggressiveness)
pred_tta = pred_tta[:, :, roi_size // 2:]
pred_tta = pred_tta[:, :, :n_frame]
return (pred + pred_tta) * 0.5 * coef, X_mag, np.exp(1.j * X_phase)
else:
return pred * coef, X_mag, np.exp(1.j * X_phase)
aggressiveness = {'value': args.aggressiveness, 'split_bin': mp.param['band'][1]['crop_stop']}
if data['tta']:
text_widget.write(base_text + "Running Inferences (TTA)...\n")
else:
text_widget.write(base_text + "Running Inference...\n")
pred, X_mag, X_phase = inference(X_spec_m,
device,
model, aggressiveness)
text_widget.write(base_text + 'Done!\n') text_widget.write(base_text + 'Done!\n')
update_progress(**progress_kwargs, update_progress(**progress_kwargs,
step=0.6) step=0.9)
# Postprocess # Postprocess
if data['postprocess']: if data['postprocess']:
text_widget.write(base_text + 'Post processing...\n') text_widget.write(base_text + 'Post processing...\n')
@@ -375,7 +385,7 @@ def main(window: tk.Wm, text_widget: tk.Text, button_widget: tk.Button, progress
text_widget.write(base_text + 'Done!\n') text_widget.write(base_text + 'Done!\n')
update_progress(**progress_kwargs, update_progress(**progress_kwargs,
step=0.65) step=0.95)
# Inverse stft # Inverse stft
text_widget.write(base_text + 'Inverse stft of instruments and vocals...\n') # nopep8 text_widget.write(base_text + 'Inverse stft of instruments and vocals...\n') # nopep8