mirror of
https://github.com/macaodha/batdetect2.git
synced 2026-04-04 15:20:19 +02:00
1.2 KiB
1.2 KiB
How to configure spectrogram preprocessing
Use this guide to set STFT, frequency range, and spectrogram transforms.
1) Configure STFT and frequency range
preprocess:
stft:
window_duration: 0.002
window_overlap: 0.75
window_fn: hann
frequencies:
min_freq: 10000
max_freq: 120000
2) Configure spectrogram transforms
spectrogram_transforms are applied in order.
preprocess:
spectrogram_transforms:
- name: pcen
time_constant: 0.4
gain: 0.98
bias: 2.0
power: 0.5
- name: spectral_mean_subtraction
- name: scale_amplitude
scale: db
Common built-ins:
pcenspectral_mean_subtractionscale_amplitude(dborpower)peak_normalize
3) Configure output size
preprocess:
size:
height: 128
resize_factor: 0.5
4) Keep train and inference settings aligned
Use the same preprocessing setup for training and prediction whenever possible. Large mismatches can degrade model performance.
Related pages
- Why consistency matters: {doc}
../explanation/preprocessing-consistency - Preprocessing config reference: {doc}
../reference/preprocessing-config