Orcasound's network, new nodes and app

15 years of citizen scientists listening for whales in the Salish Sea

Orcasound hackathon challenges

• Build new HTML5 version of orca call catalog (old version)
  1. Use ffmpeg to convert audio files (flac -> ogg + mp3)
  2. Use Python to generate spectrograms from audio data (ideally flac)
• Build training data set of killer whale signals for machine learning
  1. Listen to expert's catalog of signals (~40 types of "pulsed calls")
  2. Clip isolated, clear calls from continuous recordings of wild killer whales
• Discuss and implement architecture for machine learning
  1. Review existing data architecture and flow
  2. Adjust AWS security and set up hackathon participant access
  3. Access FLAC audio data in AWS S3 buckets
  4. Discuss and access EC2 Linux instance and other relevant services (e.g. Cloud9, Lambda?)
  5. Build out Orcadata Git repository: a docker-based Python analysis environment, or ??
• Integrate machine learning approach and products into Orcasound app
  1. Orientation to the app, audio data scheme, and Orcasite Git repository (Backend is an Elixir app using the Phoenix framework. The Phoenix app serves a React app.)
  2. Discuss how algorithmic and human detection (Activity button presses) and classification could work synergistically
  3. Implement integration, including visualization and notifications (moderated or automated)