Bioacoustics and Machine Learning as Key Tools in Conservation
summary
Conservation Metrics describes processing massive acoustic datasets — 26 terabytes, 21 million files from 1,600 locations — using cloud storage, a web labeling app, and convolutional neural networks to turn 2‑second windows into call‑per‑minute metrics for trend models. Examples: seabird collision detection drives line burial and deterrents; Kauai call rates rebound; coral reef work finds distinct parrotfish crunch and unknown “knock chorus” classes; models work unevenly across sensor types and still need more cross‑recorder training.
AI-written summary · about
chapters (17)
- 0:00Bioacoustics and conservation intro
- 1:13Conservation Metrics in acoustic monitoring
- 3:01Deep learning for conservation acoustics
- 4:50Seabird behavior with acoustic data
- 5:54ML for continental-scale bat monitoring
- 7:20Cloud tools for acoustic data processing
- 9:11Marine soundscape with coral reef models
- 11:58Unknown fish sounds in Moorea research
- 13:45Parrotfish sounds and reef health
- 15:13Fish knock chorus patterns
- 16:40Labeling challenges in marine acoustics
- 18:50Future of marine acoustic pipelines
- 19:17Q&A on sensors and data
- 22:13Data sharing in conservation
- 24:31Interference from fishfinders and sonars
- 25:55Open data challenges in acoustics
- 27:20Q&A on fish vocalizations
related talks
39mOverview and Marine Bioacoustics and Perch Model
15mThe Internet of Wild Things: Acoustics, AI and Conservation
11mAI and Bioacoustics: Tracking Biodiversity Progress with BirdNET
25mThe Sound of Life
17mTechnology: A Changemaker in Environmental Protection
28mConservation & Tech
50mRestoring Coral Reefs at Scale: AI and the Next Generation of Conservation
43mDeciphering animal emotions using machine learning and vocalisations