intial changes for V5 challenge
This commit is contained in:
Hari Dubey
Родитель
5582dcf5ba
Коммит
b5e1e13685
|
|
@ -0,0 +1,136 @@
|
|||
#!/usr/bin/bash
|
||||
|
||||
# ***** Datasets for ICASSP 2022 DNS Challenge 4 - Personalized DNS Track *****
|
||||
|
||||
# NOTE: Before downloading, make sure you have enough space
|
||||
# on your local storage!
|
||||
|
||||
# In all, you will need about 380TB to store the UNPACKED data.
|
||||
# Archived, the same data takes about 200GB total.
|
||||
|
||||
# Please comment out the files you don't need before launching
|
||||
# the script.
|
||||
|
||||
# NOTE: By default, the script *DOES NOT* DOWNLOAD ANY FILES!
|
||||
# Please scroll down and edit this script to pick the
|
||||
# downloading method that works best for you.
|
||||
|
||||
# -------------------------------------------------------------
|
||||
# The directory structure of the unpacked data is:
|
||||
|
||||
# . 362G
|
||||
# +-- datasets_fullband 64G
|
||||
# | +-- impulse_responses 5.9G
|
||||
# | \-- noise_fullband 58G
|
||||
# +-- pdns_training_set 294G
|
||||
# | +-- enrollment_embeddings 115M
|
||||
# | +-- enrollment_wav 42G
|
||||
# | +-- raw/clean 252G
|
||||
# | +-- english 168G
|
||||
# | +-- french 2.1G
|
||||
# | +-- german 53G
|
||||
# | +-- italian 17G
|
||||
# | +-- russian 6.8G
|
||||
# | \-- spanish 5.4G
|
||||
# \-- personalized_dev_testset 3.3G
|
||||
|
||||
$BLOB_NAMES=@(
|
||||
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_000.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_001.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_002.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_003.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_004.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_005.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_006.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_007.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_008.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_009.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_010.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_011.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_012.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_013.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_014.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_015.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_016.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_017.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_018.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_019.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.english_020.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.french_000.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_000.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_001.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_002.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_003.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_004.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_005.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_006.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_007.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.german_008.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.italian_000.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.italian_001.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.italian_002.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.russian_000.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.spanish_000.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.spanish_001.tar.bz2",
|
||||
"pdns_training_set/raw/pdns_training_set.raw.clean.spanish_002.tar.bz2",
|
||||
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_000.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_001.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_002.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_003.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_004.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.french_000.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.german_000.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.german_001.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.italian_000.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.russian_000.tar.bz2",
|
||||
"pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.spanish_000.tar.bz2",
|
||||
|
||||
"pdns_training_set/pdns_training_set.enrollment_embeddings_000.tar.bz2",
|
||||
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_000.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_001.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_002.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_003.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_004.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_005.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_006.tar.bz2",
|
||||
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.freesound_000.tar.bz2",
|
||||
"datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.freesound_001.tar.bz2",
|
||||
|
||||
"datasets_fullband/datasets_fullband.impulse_responses_000.tar.bz2",
|
||||
|
||||
"personalized_dev_testset/personalized_dev_testset.enrollment.tar.bz2",
|
||||
"personalized_dev_testset/personalized_dev_testset.noisy_testclips.tar.bz2"
|
||||
)
|
||||
|
||||
###############################################################
|
||||
|
||||
$AZURE_URL="https://dns4public.blob.core.windows.net/dns4archive"
|
||||
|
||||
$OUTPUT_PATH="."
|
||||
|
||||
mkdir -Force $OUTPUT_PATH/pdns_training_set/raw 2> $null
|
||||
mkdir -Force $OUTPUT_PATH/pdns_training_set/enrollment_wav 2> $null
|
||||
mkdir -Force $OUTPUT_PATH/datasets_fullband/noise_fullband 2> $null
|
||||
|
||||
foreach ($BLOB in $BLOB_NAMES) {
|
||||
|
||||
$URL="$AZURE_URL/$BLOB"
|
||||
Write-Output "Download: $BLOB"
|
||||
|
||||
# DRY RUN: print HTTP response and Content-Length
|
||||
# WITHOUT downloading the files
|
||||
curl -s -I "$URL" | head -n 2
|
||||
|
||||
# Actually download the files: UNCOMMENT when ready to download
|
||||
# curl "$URL" -o "$OUTPUT_PATH/$BLOB"
|
||||
|
||||
# Same as above, but using wget
|
||||
# wget "$URL" -O "$OUTPUT_PATH/$BLOB"
|
||||
|
||||
# Same, + unpack files on the fly
|
||||
# curl "$URL" | tar -C "$OUTPUT_PATH" -f - -x -j
|
||||
}
|
||||
|
|
@ -0,0 +1,134 @@
|
|||
#!/usr/bin/bash
|
||||
|
||||
# ***** Datasets for ICASSP 2022 DNS Challenge 4 - Personalized DNS Track *****
|
||||
|
||||
# NOTE: Before downloading, make sure you have enough space
|
||||
# on your local storage!
|
||||
|
||||
# In all, you will need about 380TB to store the UNPACKED data.
|
||||
# Archived, the same data takes about 200GB total.
|
||||
|
||||
# Please comment out the files you don't need before launching
|
||||
# the script.
|
||||
|
||||
# NOTE: By default, the script *DOES NOT* DOWNLOAD ANY FILES!
|
||||
# Please scroll down and edit this script to pick the
|
||||
# downloading method that works best for you.
|
||||
|
||||
# -------------------------------------------------------------
|
||||
# The directory structure of the unpacked data is:
|
||||
|
||||
# . 362G
|
||||
# +-- datasets_fullband 64G
|
||||
# | +-- impulse_responses 5.9G
|
||||
# | \-- noise_fullband 58G
|
||||
# +-- pdns_training_set 294G
|
||||
# | +-- enrollment_embeddings 115M
|
||||
# | +-- enrollment_wav 42G
|
||||
# | +-- raw/clean 252G
|
||||
# | +-- english 168G
|
||||
# | +-- french 2.1G
|
||||
# | +-- german 53G
|
||||
# | +-- italian 17G
|
||||
# | +-- russian 6.8G
|
||||
# | \-- spanish 5.4G
|
||||
# \-- personalized_dev_testset 3.3G
|
||||
|
||||
BLOB_NAMES=(
|
||||
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_000.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_001.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_002.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_003.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_004.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_005.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_006.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_007.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_008.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_009.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_010.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_011.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_012.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_013.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_014.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_015.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_016.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_017.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_018.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_019.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.english_020.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.french_000.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_000.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_001.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_002.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_003.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_004.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_005.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_006.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_007.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.german_008.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.italian_000.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.italian_001.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.italian_002.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.russian_000.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.spanish_000.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.spanish_001.tar.bz2
|
||||
pdns_training_set/raw/pdns_training_set.raw.clean.spanish_002.tar.bz2
|
||||
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_000.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_001.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_002.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_003.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.english_004.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.french_000.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.german_000.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.german_001.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.italian_000.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.russian_000.tar.bz2
|
||||
pdns_training_set/enrollment_wav/pdns_training_set.enrollment_wav.spanish_000.tar.bz2
|
||||
|
||||
pdns_training_set/pdns_training_set.enrollment_embeddings_000.tar.bz2
|
||||
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_000.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_001.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_002.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_003.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_004.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_005.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.audioset_006.tar.bz2
|
||||
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.freesound_000.tar.bz2
|
||||
datasets_fullband/noise_fullband/datasets_fullband.noise_fullband.freesound_001.tar.bz2
|
||||
|
||||
datasets_fullband/datasets_fullband.impulse_responses_000.tar.bz2
|
||||
|
||||
personalized_dev_testset/personalized_dev_testset.enrollment.tar.bz2
|
||||
personalized_dev_testset/personalized_dev_testset.noisy_testclips.tar.bz2
|
||||
)
|
||||
|
||||
###############################################################
|
||||
|
||||
AZURE_URL="https://dns4public.blob.core.windows.net/dns4archive"
|
||||
|
||||
OUTPUT_PATH="."
|
||||
|
||||
mkdir -p $OUTPUT_PATH/{pdns_training_set/{raw,enrollment_wav},datasets_fullband/noise_fullband}
|
||||
|
||||
for BLOB in ${BLOB_NAMES[@]}
|
||||
do
|
||||
URL="$AZURE_URL/$BLOB"
|
||||
echo "Download: $BLOB"
|
||||
|
||||
# DRY RUN: print HTTP response and Content-Length
|
||||
# WITHOUT downloading the files
|
||||
curl -s -I "$URL" | head -n 2
|
||||
|
||||
# Actually download the files: UNCOMMENT when ready to download
|
||||
# curl "$URL" -o "$OUTPUT_PATH/$BLOB"
|
||||
|
||||
# Same as above, but using wget
|
||||
# wget "$URL" -O "$OUTPUT_PATH/$BLOB"
|
||||
|
||||
# Same, + unpack files on the fly
|
||||
# curl "$URL" | tar -C "$OUTPUT_PATH" -f - -x -j
|
||||
done
|
||||
|
|
@ -1,137 +1,137 @@
|
|||
# Configuration for generating Noisy Speech Dataset
|
||||
|
||||
# - sampling_rate: Specify the sampling rate. Default is 16 kHz
|
||||
# - audioformat: default is .wav
|
||||
# - audio_length: Minimum Length of each audio clip (noisy and clean speech) in seconds that will be generated by augmenting utterances.
|
||||
# - silence_length: Duration of silence introduced between clean speech utterances.
|
||||
# - total_hours: Total number of hours of data required. Units are in hours.
|
||||
# - snr_lower: Lower bound for SNR required (default: 0 dB)
|
||||
# - snr_upper: Upper bound for SNR required (default: 40 dB)
|
||||
# - target_level_lower: Lower bound for the target audio level before audiowrite (default: -35 dB)
|
||||
# - target_level_upper: Upper bound for the target audio level before audiowrite (default: -15 dB)
|
||||
# - total_snrlevels: Number of SNR levels required (default: 5, which means there are 5 levels between snr_lower and snr_upper)
|
||||
# - clean_activity_threshold: Activity threshold for clean speech
|
||||
# - noise_activity_threshold: Activity threshold for noise
|
||||
# - fileindex_start: Starting file ID that will be used in filenames
|
||||
# - fileindex_end: Last file ID that will be used in filenames
|
||||
# - is_test_set: Set it to True if it is the test set, else False for the training set
|
||||
# - noise_dir: Specify the directory path to all noise files
|
||||
# - Speech_dir: Specify the directory path to all clean speech files
|
||||
# - noisy_destination: Specify path to the destination directory to store noisy speech
|
||||
# - clean_destination: Specify path to the destination directory to store clean speech
|
||||
# - noise_destination: Specify path to the destination directory to store noise speech
|
||||
# - log_dir: Specify path to the directory to store all the log files
|
||||
|
||||
# Configuration for unit tests
|
||||
# - snr_test: Set to True if SNR test is required, else False
|
||||
# - norm_test: Set to True if Normalization test is required, else False
|
||||
# - sampling_rate_test: Set to True if Sampling Rate test is required, else False
|
||||
# - clipping_test: Set to True if Clipping test is required, else False
|
||||
# - unit_tests_log_dir: Specify path to the directory where you want to store logs
|
||||
|
||||
[noisy_speech]
|
||||
|
||||
sampling_rate: 48000
|
||||
audioformat: *.wav
|
||||
audio_length: 30
|
||||
# 15, 12, 30
|
||||
silence_length: 0.2
|
||||
total_hours: 1000
|
||||
# 1000
|
||||
#200
|
||||
# 2.5, 500, 100
|
||||
snr_lower: -5
|
||||
#-5, 0
|
||||
snr_upper: 20
|
||||
# 25, 40
|
||||
randomize_snr: True
|
||||
target_level_lower: -35
|
||||
target_level_upper: -15
|
||||
total_snrlevels: 31
|
||||
# 5
|
||||
clean_activity_threshold: 0.0
|
||||
noise_activity_threshold: 0.2
|
||||
fileindex_start: None
|
||||
fileindex_end: None
|
||||
is_test_set: False
|
||||
# True, False
|
||||
|
||||
noise_dir: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/noise
|
||||
#/mnt/f/4th_DNSChallenge/INTERSPEECH_2021/DNS-Challenge/datasets_fullband/noise
|
||||
#F:\4th_DNSChallenge\INTERSPEECH_2021\DNS-Challenge\datasets_fullband\noise
|
||||
#datasets\pdns_training_set\noise
|
||||
#\test_set2\Test_Noise
|
||||
# datasets\noise
|
||||
# \datasets\noise
|
||||
|
||||
speech_dir: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/clean
|
||||
# D:\kanhawin_git\primary_speakers_VCTK_16k_for_synthesizer
|
||||
# datasets\test_set2\Singing_Voice\wav_16k
|
||||
# dir with secondary speaker clean speech
|
||||
speech_dir2: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/clean
|
||||
#D:\kanhawin_git\secondary_speakers_voxCeleb2_16k
|
||||
# datasets\test_set2\Singing_Voice\wav_16k
|
||||
|
||||
spkid_csv: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/filelists/complete_ps_split.csv
|
||||
#/mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/filelists/vctk_spkid.csv
|
||||
# datasets\clean
|
||||
noise_types_excluded: None
|
||||
|
||||
rir_dir: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/impulse_responses
|
||||
#/mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/impulse_responses
|
||||
# F:\4th_DNSChallenge\ICASSP_2022\DNS-Challenge\datasets\impulse_responses
|
||||
|
||||
# \datasets\clean
|
||||
noisy_destination: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/mixed/noisy
|
||||
# datasets/training_data/noisy
|
||||
# datasets\test_set2\synthetic_personalizeddns\noisy
|
||||
#training_set2_onlyrealrir\noisy
|
||||
#\noisy
|
||||
clean_destination: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/mixed/clean
|
||||
#datasets\test_set2\synthetic_personalizeddns\clean
|
||||
# training_set2_onlyrealrir\clean
|
||||
# \clean
|
||||
noise_destination: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/mixed/noise
|
||||
# datasets/training_data/noise
|
||||
#datasets\test_set2\synthetic_personalizeddns\noise
|
||||
#training_set2_onlyrealrir\noise
|
||||
# \noise
|
||||
log_dir: logs
|
||||
# \logs
|
||||
|
||||
# Config: add singing voice to clean speech
|
||||
clean_singing: datasets\clean_singing\VocalSet11\FULL
|
||||
singing_choice: 3
|
||||
# 1 for only male, 2 for only female, 3 (default) for both male and female
|
||||
|
||||
# Config: add reverb to clean speech
|
||||
rir_choice: 1
|
||||
# 1 for only real rir, 2 for only synthetic rir, 3 (default) use both real and synthetic
|
||||
lower_t60: 0.3
|
||||
# lower bound of t60 range in seconds
|
||||
upper_t60: 1.3
|
||||
# upper bound of t60 range in seconds
|
||||
rir_table_csv: datasets\acoustic_params\RIR_table_simple.csv
|
||||
clean_speech_t60_csv: datasets\acoustic_params\cleanspeech_table_t60_c50.csv
|
||||
# percent_for_adding_reverb=0.5 # percentage of clean speech convolved with RIR
|
||||
|
||||
# pdns testsets
|
||||
# primary_data: D:\kanhawin_git\primary_speakers_VCTK_16k
|
||||
#'D:\PersonalizedDNS_dataset\synthetic_primary'
|
||||
# secondary_data='D:\kanhawin_git\secondary_speakers_voxCeleb2_16k'
|
||||
#'D:\PersonalizedDNS_dataset\synthetic_secondary'
|
||||
# noise_data= datasets\test_set2\synthetic\noise
|
||||
# pdns_testset_clean= datasets\test_set2\pdns\clean
|
||||
# pdns_testset_noisy= datasets\test_set2\pdns\noisy
|
||||
|
||||
# adaptation_data_seconds=120
|
||||
# num_primary_spk=100
|
||||
# num_clips=600
|
||||
|
||||
# Unit tests config
|
||||
snr_test: True
|
||||
norm_test: True
|
||||
sampling_rate_test = True
|
||||
clipping_test = True
|
||||
|
||||
unit_tests_log_dir: unittests_logs
|
||||
# Configuration for generating Noisy Speech Dataset
|
||||
|
||||
# - sampling_rate: Specify the sampling rate. Default is 16 kHz
|
||||
# - audioformat: default is .wav
|
||||
# - audio_length: Minimum Length of each audio clip (noisy and clean speech) in seconds that will be generated by augmenting utterances.
|
||||
# - silence_length: Duration of silence introduced between clean speech utterances.
|
||||
# - total_hours: Total number of hours of data required. Units are in hours.
|
||||
# - snr_lower: Lower bound for SNR required (default: 0 dB)
|
||||
# - snr_upper: Upper bound for SNR required (default: 40 dB)
|
||||
# - target_level_lower: Lower bound for the target audio level before audiowrite (default: -35 dB)
|
||||
# - target_level_upper: Upper bound for the target audio level before audiowrite (default: -15 dB)
|
||||
# - total_snrlevels: Number of SNR levels required (default: 5, which means there are 5 levels between snr_lower and snr_upper)
|
||||
# - clean_activity_threshold: Activity threshold for clean speech
|
||||
# - noise_activity_threshold: Activity threshold for noise
|
||||
# - fileindex_start: Starting file ID that will be used in filenames
|
||||
# - fileindex_end: Last file ID that will be used in filenames
|
||||
# - is_test_set: Set it to True if it is the test set, else False for the training set
|
||||
# - noise_dir: Specify the directory path to all noise files
|
||||
# - Speech_dir: Specify the directory path to all clean speech files
|
||||
# - noisy_destination: Specify path to the destination directory to store noisy speech
|
||||
# - clean_destination: Specify path to the destination directory to store clean speech
|
||||
# - noise_destination: Specify path to the destination directory to store noise speech
|
||||
# - log_dir: Specify path to the directory to store all the log files
|
||||
|
||||
# Configuration for unit tests
|
||||
# - snr_test: Set to True if SNR test is required, else False
|
||||
# - norm_test: Set to True if Normalization test is required, else False
|
||||
# - sampling_rate_test: Set to True if Sampling Rate test is required, else False
|
||||
# - clipping_test: Set to True if Clipping test is required, else False
|
||||
# - unit_tests_log_dir: Specify path to the directory where you want to store logs
|
||||
|
||||
[noisy_speech]
|
||||
|
||||
sampling_rate: 48000
|
||||
audioformat: *.wav
|
||||
audio_length: 30
|
||||
# 15, 12, 30
|
||||
silence_length: 0.2
|
||||
total_hours: 1000
|
||||
# 1000
|
||||
#200
|
||||
# 2.5, 500, 100
|
||||
snr_lower: -5
|
||||
#-5, 0
|
||||
snr_upper: 20
|
||||
# 25, 40
|
||||
randomize_snr: True
|
||||
target_level_lower: -35
|
||||
target_level_upper: -15
|
||||
total_snrlevels: 31
|
||||
# 5
|
||||
clean_activity_threshold: 0.0
|
||||
noise_activity_threshold: 0.2
|
||||
fileindex_start: None
|
||||
fileindex_end: None
|
||||
is_test_set: False
|
||||
# True, False
|
||||
|
||||
noise_dir: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/noise
|
||||
#/mnt/f/4th_DNSChallenge/INTERSPEECH_2021/DNS-Challenge/datasets_fullband/noise
|
||||
#F:\4th_DNSChallenge\INTERSPEECH_2021\DNS-Challenge\datasets_fullband\noise
|
||||
#datasets\pdns_training_set\noise
|
||||
#\test_set2\Test_Noise
|
||||
# datasets\noise
|
||||
# \datasets\noise
|
||||
|
||||
speech_dir: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/clean
|
||||
# D:\kanhawin_git\primary_speakers_VCTK_16k_for_synthesizer
|
||||
# datasets\test_set2\Singing_Voice\wav_16k
|
||||
# dir with secondary speaker clean speech
|
||||
speech_dir2: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/clean
|
||||
#D:\kanhawin_git\secondary_speakers_voxCeleb2_16k
|
||||
# datasets\test_set2\Singing_Voice\wav_16k
|
||||
|
||||
spkid_csv: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/filelists/complete_ps_split.csv
|
||||
#/mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/filelists/vctk_spkid.csv
|
||||
# datasets\clean
|
||||
noise_types_excluded: None
|
||||
|
||||
rir_dir: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/impulse_responses
|
||||
#/mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/impulse_responses
|
||||
# F:\4th_DNSChallenge\ICASSP_2022\DNS-Challenge\datasets\impulse_responses
|
||||
|
||||
# \datasets\clean
|
||||
noisy_destination: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/mixed/noisy
|
||||
# datasets/training_data/noisy
|
||||
# datasets\test_set2\synthetic_personalizeddns\noisy
|
||||
#training_set2_onlyrealrir\noisy
|
||||
#\noisy
|
||||
clean_destination: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/mixed/clean
|
||||
#datasets\test_set2\synthetic_personalizeddns\clean
|
||||
# training_set2_onlyrealrir\clean
|
||||
# \clean
|
||||
noise_destination: /mnt/f/4th_DNSChallenge/ICASSP_2022/DNS-Challenge/datasets/pdns_training_set/mixed/noise
|
||||
# datasets/training_data/noise
|
||||
#datasets\test_set2\synthetic_personalizeddns\noise
|
||||
#training_set2_onlyrealrir\noise
|
||||
# \noise
|
||||
log_dir: logs
|
||||
# \logs
|
||||
|
||||
# Config: add singing voice to clean speech
|
||||
clean_singing: datasets\clean_singing\VocalSet11\FULL
|
||||
singing_choice: 3
|
||||
# 1 for only male, 2 for only female, 3 (default) for both male and female
|
||||
|
||||
# Config: add reverb to clean speech
|
||||
rir_choice: 1
|
||||
# 1 for only real rir, 2 for only synthetic rir, 3 (default) use both real and synthetic
|
||||
lower_t60: 0.3
|
||||
# lower bound of t60 range in seconds
|
||||
upper_t60: 1.3
|
||||
# upper bound of t60 range in seconds
|
||||
rir_table_csv: datasets\acoustic_params\RIR_table_simple.csv
|
||||
clean_speech_t60_csv: datasets\acoustic_params\cleanspeech_table_t60_c50.csv
|
||||
# percent_for_adding_reverb=0.5 # percentage of clean speech convolved with RIR
|
||||
|
||||
# pdns testsets
|
||||
# primary_data: D:\kanhawin_git\primary_speakers_VCTK_16k
|
||||
#'D:\PersonalizedDNS_dataset\synthetic_primary'
|
||||
# secondary_data='D:\kanhawin_git\secondary_speakers_voxCeleb2_16k'
|
||||
#'D:\PersonalizedDNS_dataset\synthetic_secondary'
|
||||
# noise_data= datasets\test_set2\synthetic\noise
|
||||
# pdns_testset_clean= datasets\test_set2\pdns\clean
|
||||
# pdns_testset_noisy= datasets\test_set2\pdns\noisy
|
||||
|
||||
# adaptation_data_seconds=120
|
||||
# num_primary_spk=100
|
||||
# num_clips=600
|
||||
|
||||
# Unit tests config
|
||||
snr_test: True
|
||||
norm_test: True
|
||||
sampling_rate_test = True
|
||||
clipping_test = True
|
||||
|
||||
unit_tests_log_dir: unittests_logs
|
||||
Загрузка…
Ссылка в новой задаче