LOCAL_RUN.md

Local Run

You can simply run python test.py for running your submissions locally. Few default environment variables:

• TEST_DATASET_PATH (default: data/test/): path to the test dataset folder.
• RESULTS_DATASET_PATH (default: data/results/): path to the results dataset folder.
• INFERENCE_SETUP_TIMEOUT_SECONDS (default: 900 seconds): timeout for your predict_setup function.
• INFERENCE_PER_MUSIC_TIMEOUT_SECONDS (default: 240 seconds): timeout for your predict function.

python test.py

Directory structure after running will look something like:

.
├── test
│   ├── SS_008
│   │   └── mixture.wav
│   └── SS_018
│       └── mixture.wav
└── results
    ├── SS_008
    │   ├── bass.wav
    │   ├── drums.wav
    │   ├── mixture.wav
    │   ├── other.wav
    │   └── vocals.wav
    └── SS_018
        ├── bass.wav
        ├── drums.wav
        ├── mixture.wav
        ├── other.wav
        └── vocals.wav

Scoring (local)

You can also calculate scores for your local evaluation by running:

python score.py

This will compare files present in test/ folder with results/ folder for SDR calculation.

def sdr(references, estimates):
    # compute SDR for one song
    delta = 1e-7  # avoid numerical errors
    num = np.sum(np.square(references), axis=(1, 2))
    den = np.sum(np.square(references - estimates), axis=(1, 2))
    num += delta
    den += delta
    return 10 * np.log10(num  / den)