Particle.news
Download on the App Store
3 ARTICLES
·
last mo.

IIAP Mobilizes AI and Supercomputer to Scan 800,000 Minutes of Amazon Sound for Human Impact

Researchers are training machine‑learning models on the Manatí supercomputer to separate animal calls from human-made noise and publish a public acoustic library by year-end.

Image

Overview

  • The Institute of Research for the Peruvian Amazon (IIAP) has completed collection of more than 800,000 minutes of environmental audio from roads, settlements and protected sites across the Peruvian Amazon.
  • Teams placed automated recorders at sites including the Iquitos–Nauta, San Joaquín de Omaguas and Jenaro Herrera corridors and the Allpahuayo Mishana reserve to capture contrast between impacted and conserved soundscapes.
  • The IIAP Artificial Intelligence Laboratory is using high‑performance systems, led by the Manatí supercomputer, to train models that will process the large audio volume far faster than manual methods.
  • Researchers are building a labeled reference database so algorithms can tell biological vocalizations apart from anthropogenic noise, a key step to reliably identify species that are hard to see in the forest.
  • The project will publish a public acoustic library toward the end of the year to support researchers, teachers and local managers and to provide evidence for which species persist or disappear as human pressure increases.