This is a script I use to navigate the sound dataset from FSK50K. https://zenodo.org/record/4060432#.YYYJ9HtBxH4

To use this download the files from datasets. The categories is in the Ground truth file to be downloaded in the links above.

These have several script

• find_audio.py - This is the script you want
• convert.py - this only exist so that I can convert data for a Wio Terminal project. https://wiki.seeedstudio.com/Wio-Terminal-TinyML-EI-3/
• sound_category.py - this is to show the category of a file

• set the path to the ground truth csv path
• export export FSD_PATH="/somepath/somewhere/FSD50K.ground_truth/dev.csv"

• pipenv install
• Start the shell with pipenv shell then use the script as described below.

Usage

• Basic use - python find_audio.py categories
• Show all label of a file - python find_audio.py categories -s
• Exclude file with certain categories - python find_audio.py categories -x xcategories
• Search Multiple file - python find_audio.py cat1,cat2,etc
• Search Exclude Multiple file - python find_audio.py cat1,cat2,etc -x xcat1,xcat2
• Help - python find_audio.py -h

Usage is similar for find_audio.py except for -p, -i and -o

• Basic use - python convert.py categories -p new_category_name -i source_audio_directory -o output_directory

• Basic use - python sound_category.py audio_name