Download Audioset V2 with FFmpeg. Requires proxychains + Ffmpeg + gnu parallel

download_audioset.sh

# @Author: richman
# @Date:   2018-03-15
# @Last Modified by:   richman
# @Last Modified time: 2018-03-30

if [[ $# < 1 ]]; then
    echo "Input .csv file .e.g balanced_train_segments.csv"
    exit
fi

inp=$1
njobs=${2:-4}

SAMPLE_RATE=16000
EXTENSION="flac"

fetch_clip() {
    # echo "Fetching $1 ($2 to $3)..."
    outname="$1_$2_$3"
    if [ -f "${outname}.${EXTENSION}" ]; then
        return
    fi
    link=$(youtube-dl -g https://youtube.com/watch?v=$1 | awk 'NR==2{print}')

    if [ $? -eq 0 ]; then
        proxychains -q ffmpeg -loglevel quiet  -i "$link" -ar $SAMPLE_RATE \
        -ss "$2" -to "$3" "./${outname}.${EXTENSION}"
    fi
}


export SAMPLE_RATE
export EXTENSION
export -f fetch_clip


#until proxychains -q wget -q --spider http://google.com
#do
    #echo "No connection. Sleeping 1 minute."
    #sleep 1m;
#done

grep "^[^#;]" $inp | parallel --resume --joblog job.log -j $njobs --colsep=, fetch_clip {1} {2} {3} > /dev/null