Acoustic Data Pipeline Building and Validation

 

A comparison of two manatee calls. We are building a pipeline for data annotation, and one important step in the process is to validate previously made annotations. The quality of data is of extreme importance, and we use experts with domain knowledge to validate all annotations that will be included in our training set. Here is an example of a previously made annotation that rated two manatee calls as different with a 0% similarity score. These two files will be positively validated and will become a part of our training dataset.