Improvements to make_frames and audio class
This merge request
-
adds a warning in
make_frames
method in audio_processing module, issued when the estimated size of the output frames exceeds 10% of system memory. -
adds a
segment
method to the audio class, which splits the audio signal into equal length segments, while also handling annotations -
adds a check of memory usage in the
create_spec_database
method; if too much memory is used, the audio signal is segmented before the spectrogram is computed -
improves parsing of file names in the audio class to ensure correct behaviour also on Windows
-
adds option to enforce same length when extracting annotated segments from a spectrogram. If an annotation is shorter than the specified length, the annotation box is stretched; if it is shorter, the box is divided into several segments.
Obs: Remember to update development log when merging
Obs-obs: Remember to remove pip install psutil
from gitlab-ci.yml script once Docker image has been updated