Raven to wav is used to filter annotated parts from the original .wav files based on annotations in table format (e.g. .txt or .csv).
In condensation we use an energy change based method to filter out low energy parts from the original dataset to make manual annotation/labeling more efficient.
In Synthetic data we embed Chimpanze vocalizations in jungle sounds that are labeled as background to create more and more diverse data.
Chunk_wav component splits .wav files into smaller chunks. This is a step in the preparation process for deep
learning models. By default, desired length of wav files is 0.5 seconds. There is an overlap of 0.25 seconds between each of two
following chunks.
Find usage steps in the respective folders.