ASesYusuf1/SESA_Audio_Separation

1# Kaggle için gerekli kütüphaneleri yükleme2import os3import glob4import subprocess5import time6import gc7import shutil8import sys9from datetime import datetime10import torch11import yaml12import numpy as np13import requests14import json15import locale16import re17import psutil18import concurrent.futures19from tqdm import tqdm20from urllib.parse import urlparse, quote21 22# Ses işleme kütüphaneleri23!pip install librosa soundfile24import librosa25import soundfile as sf26 27# Uyarıları görmezden gel28import warnings29warnings.filterwarnings("ignore")30 31# Kaggle dizin ayarları32BASE_DIR = "/kaggle/working/"  # Kaggle'da çalışma dizini33INFERENCE_PATH = os.path.join(BASE_DIR, "inference.py")  # inference.py varsa buraya kopyalanmalı34OUTPUT_DIR = os.path.join(BASE_DIR, "output")  # Çıkış dizini35AUTO_ENSEMBLE_OUTPUT = os.path.join(BASE_DIR, "ensemble_output")  # Ensemble çıkış dizini36 37# Çıkış dizinlerini oluştur38os.makedirs(OUTPUT_DIR, exist_ok=True)39os.makedirs(AUTO_ENSEMBLE_OUTPUT, exist_ok=True)40 41def clean_model_name(model):42    """43    Clean and standardize model names for filename44    """45    model_name_mapping = {46        'VOCALS-InstVocHQ': 'InstVocHQ',47        'VOCALS-MelBand-Roformer (by KimberleyJSN)': 'KimberleyJSN',48        'VOCALS-BS-Roformer_1297 (by viperx)': 'VOCALS_BS_Roformer1297',49        'VOCALS-BS-Roformer_1296 (by viperx)': 'VOCALS-BS-Roformer_1296',50        'VOCALS-BS-RoformerLargev1 (by unwa)': 'UnwaLargeV1',51        'VOCALS-Mel-Roformer big beta 4 (by unwa)': 'UnwaBigBeta4',52        'VOCALS-Melband-Roformer BigBeta5e (by unwa)': 'UnwaBigBeta5e',53        'INST-Mel-Roformer v1 (by unwa)': 'UnwaInstV1',54        'INST-Mel-Roformer v2 (by unwa)': 'UnwaInstV2',55        'INST-VOC-Mel-Roformer a.k.a. duality (by unwa)': 'UnwaDualityV1',56        'INST-VOC-Mel-Roformer a.k.a. duality v2 (by unwa)': 'UnwaDualityV2',57        'KARAOKE-MelBand-Roformer (by aufr33 & viperx)': 'KaraokeMelBandRoformer',58        'VOCALS-VitLarge23 (by ZFTurbo)': 'VitLarge23',59        'VOCALS-MelBand-Roformer (by Becruily)': 'BecruilyVocals',60        'INST-MelBand-Roformer (by Becruily)': 'BecruilyInst',61        'VOCALS-MelBand-Roformer Kim FT (by Unwa)': 'KimFT',62        'INST-MelBand-Roformer Kim FT (by Unwa)': 'KimFTInst',63        'OTHER-BS-Roformer_1053 (by viperx)': 'OtherViperx1053',64        'CROWD-REMOVAL-MelBand-Roformer (by aufr33)': 'CrowdRemovalRoformer',65        'CINEMATIC-BandIt_Plus (by kwatcharasupat)': 'CinematicBandItPlus',66        'DRUMSEP-MDX23C_DrumSep_6stem (by aufr33 & jarredou)': 'DrumSepMDX23C',67        '4STEMS-SCNet_MUSDB18 (by starrytong)': 'FourStemsSCNet',68        'DE-REVERB-MDX23C (by aufr33 & jarredou)': 'DeReverbMDX23C',69        'DENOISE-MelBand-Roformer-1 (by aufr33)': 'DenoiseMelBand1',70        'DENOISE-MelBand-Roformer-2 (by aufr33)': 'DenoiseMelBand2',71        'INST-MelBand-Roformer (by Becruily)': 'BecruilyInst',72        '4STEMS-SCNet_XL_MUSDB18 (by ZFTurbo)': 'FourStemsSCNetXL',73        '4STEMS-SCNet_Large (by starrytong)': 'FourStemsSCNetLarge',74        '4STEMS-BS-Roformer_MUSDB18 (by ZFTurbo)': 'FourStemsBSRoformer',75        'DE-REVERB-MelBand-Roformer aggr./v2/19.1729 (by anvuew)': 'DeReverbMelBandAggr',76        'DE-REVERB-Echo-MelBand-Roformer (by Sucial)': 'DeReverbEchoMelBand',77        'bleed_suppressor_v1 (by unwa)': 'BleedSuppressorV1',78        'inst_v1e (by unwa)': 'InstV1E',79        'inst_gabox (by Gabox)': 'InstGabox',80        'inst_gaboxBV1 (by Gabox)': 'InstGaboxBV1',81        'inst_gaboxBV2 (by Gabox)': 'InstGaboxBV2',82        'inst_gaboxBFV1 (by Gabox)': 'InstGaboxBFV1',83        'inst_gaboxFV2 (by Gabox)': 'InstGaboxFV2',84        'inst_gaboxFV1 (by Gabox)': 'InstGaboxFV1',85        'dereverb_mel_band_roformer_less_aggressive_anvuew': 'DereverbMelBandRoformerLessAggressive',86        'dereverb_mel_band_roformer_anvuew': 'DereverbMelBandRoformer',87        'VOCALS-Male Female-BS-RoFormer Male Female Beta 7_2889 (by aufr33)': 'MaleFemale-BS-RoFormer-(by aufr33)',88        'VOCALS-MelBand-Roformer (by Becruily)': 'Vocals-MelBand-Roformer-(by Becruily)',89        'VOCALS-MelBand-Roformer Kim FT 2 (by Unwa)': 'Vocals-MelBand-Roformer-KİM-FT-2(by Unwa)',90        'voc_gaboxMelRoformer (by Gabox)': 'voc_gaboxMelRoformer',91        'voc_gaboxBSroformer (by Gabox)': 'voc_gaboxBSroformer',92        'voc_gaboxMelRoformerFV1 (by Gabox)': 'voc_gaboxMelRoformerFV1',93        'voc_gaboxMelRoformerFV2 (by Gabox)': 'voc_gaboxMelRoformerFV2',94        'SYH99999/MelBandRoformerSYHFTB1(by Amane)': 'MelBandRoformerSYHFTB1',95        'inst_V5 (by Gabox)': 'INSTV5-(by Gabox)',96        'inst_Fv4Noise (by Gabox)': 'Inst_Fv4Noise-(by Gabox)',97        'Intrumental_Gabox (by Gabox)': 'Intrumental_Gabox-(by Gabox)',98        'inst_GaboxFv3 (by Gabox)': 'INST_GaboxFv3-(by Gabox)',99        'SYH99999/MelBandRoformerSYHFTB1_Model1 (by Amane)': 'MelBandRoformerSYHFTB1_model1',100        'SYH99999/MelBandRoformerSYHFTB1_Model2 (by Amane)': 'MelBandRoformerSYHFTB1_model2',101        'SYH99999/MelBandRoformerSYHFTB1_Model3 (by Amane)': 'MelBandRoformerSYHFTB1_model3',102        'VOCALS-MelBand-Roformer Kim FT 2 Blendless (by unwa)': 'VOCALS-MelBand-Roformer-Kim-FT-2-Blendless-(by unwa)',103        'inst_gaboxFV6 (by Gabox)': 'inst_gaboxFV6-(by Gabox)',104        'denoisedebleed (by Gabox)': 'denoisedebleed-(by Gabox)',105        'INSTV5N (by Gabox)': 'INSTV5N_(by Gabox)',106        'Voc_Fv3 (by Gabox)': 'Voc_Fv3_(by Gabox)',107        'MelBandRoformer4StemFTLarge (SYH99999)': 'MelBandRoformer4StemFTLarge_(SYH99999)',108        'dereverb_mel_band_roformer_mono (by anvuew)': 'dereverb_mel_band_roformer_mono_(by anvuew)',109        'INSTV6N (by Gabox)': 'INSTV6N_(by Gabox)',110        'KaraokeGabox': 'KaraokeGabox',111        'FullnessVocalModel (by Amane)': 'FullnessVocalModel',112        'Inst_GaboxV7 (by Gabox)': 'Inst_GaboxV7_(by Gabox)',113    }114 115    if model in model_name_mapping:116        return model_name_mapping[model]117    118    cleaned = re.sub(r'\s*\(.*?\)', '', model)  # Remove parenthetical info119    cleaned = cleaned.replace('-', '_')120    cleaned = ''.join(char for char in cleaned if char.isalnum() or char == '_')121    122    return cleaned123 124def shorten_filename(filename, max_length=30):125    """126    Shortens a filename to a specified maximum length127    """128    base, ext = os.path.splitext(filename)129    if len(base) <= max_length:130        return filename131    shortened = base[:15] + "..." + base[-10:] + ext132    return shortened133 134def clean_filename(filename):135    """136    Temizlenmiş dosya adını döndürür137    """138    cleanup_patterns = [139        r'_\d{8}_\d{6}_\d{6}$',  # _20231215_123456_123456140        r'_\d{14}$',              # _20231215123456141        r'_\d{10}$',              # _1702658400142        r'_\d+$'                  # Herhangi bir sayı143    ]144    145    base, ext = os.path.splitext(filename)146    for pattern in cleanup_patterns:147        base = re.sub(pattern, '', base)148    149    file_types = ['vocals', 'instrumental', 'drum', 'bass', 'other', 'effects', 'speech', 'music', 'dry', 'male', 'female']150    for type_keyword in file_types:151        base = base.replace(f'_{type_keyword}', '')152    153    detected_type = None154    for type_keyword in file_types:155        if type_keyword in base.lower():156            detected_type = type_keyword157            break158    159    clean_base = base.strip('_- ')160    return clean_base, detected_type, ext161 162# Örnek kullanım (Kaggle veri setinden dosya okuma)163# Veri setiniz /kaggle/input/ altında olmalı164# Örnek: input_file = "/kaggle/input/your-dataset/sample_audio.wav"165# clean_base, detected_type, ext = clean_filename("example_vocals_20231215.wav")166# print(clean_base, detected_type, ext)167