audioset_raw

Download and create a tfreader for the audioset dataset

Download dataset (takes few hours)

cd download/train
cat ../balanced_train_segments.csv | ../download.sh
cd ../validate 
cat ../eval_segments.csv | ../download.sh

Access the data (either through TFRecord or a Generator)

• Build the TFRecord file

cd download
python audioset_writer.py train
python audioset_writer.py validate

• Access WAV through a simple generator

from audioset_generator import get_batchs

# get_batchs is a generator
my_gen = get_batchs(batch_size=10, n_epochs=1, num_threads=4)
my_gen.next()