Skip to content

Multispecies Whale Detection

Detect and classify whale sounds across species

Visit official site
Category AI models

What Multispecies Whale Detection is

The project is described as a Google open-source toolkit that classifies whale sounds by species and region with neural networks.

What you can do with it

  • Build training inputs from labelled audio
  • Split and resample multi-channel files
  • Cut large files into short clips

Data pipeline

  • Labels and metadata are merged together
  • Output is represented as TensorFlow Examples
  • Clip start times are adjusted for labels

Pros & cons

✓ What we like

  • Several audio file formats are supported
  • Large files are cut into short training clips

! What to watch out for

  • Work centres on labelled audio collections
  • The pipeline is built around Apache Beam

FAQ

What does the pipeline create?

Training inputs from a collection of labelled audio files.

How are labels handled?

Label start times are adjusted relative to each audio clip.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review