MACS.md

MACS

Data Collection

Source	Collecting Method
MACS Zenodo page	in MACS Zenodo page, download MACS.yaml, find TAU Urban Acoustic Scenes 2019, Development dataset and download it.

Preprocessing Principles

You may refer to preprocess_MACS_yuchen.py for all the details. Here we just offer a concise summary:

Overview

Some audio-json pairs selected from the processed dataset:

1.mov

{
    "text": [
        "ball game with the ball bouncing and people running",
        "some adults are talking",
        "adults talking with an unrecognizable sound in the background"
    ],
    "tag": [
        "adults talking",
        "footsteps"
    ],
    "original_data": {
        "title": "MACS",
        "filename": "airport-stockholm-12-488-a.wav",
        "annotator_ids": [
            354,
            353,
            272
        ]
    }
}

2.mov

{
    "text": [
        "motorcycle growling people talking someone coughs",
        "footsteps followed by some alarm and a person coughing while chatting",
        "people are talking traffic noise"
    ],
    "tag": [
        "traffic noise",
        "adults talking",
        "siren",
        "footsteps"
    ],
    "original_data": {
        "title": "MACS",
        "filename": "public_square-helsinki-111-3230-a.wav",
        "annotator_ids": [
            197,
            262,
            366
        ]
    }
}

I. Json file generation principles

We use information retrieved from MACS.yaml.

• text entry For each audio, in MACS.yaml, we can find 4-5 sentences offered respectively by different annotators, we collect them and put them here as a list of string.
• tag entry Each annotator also provides a list of tags for each audio, we collect them and put them here as a list of string. (no duplicates)
• original_data entry annotator_ids(a list of annotators' ids), filename(the name of the audio file in TAU2019 dataset) are put here.

II. Audio filtering principles

Discard all audios failed to be read by soundfile.read() method or denied by FFmpeg while processing.

III. Audio format specifications

After the preprocessing work, all audio files should be in FLAC format with sampling rate of 48KHZ. (Processed by ffmpeg).