Fifty-eight of 84 known Hawaiian forest bird species (manu) have gone extinct, and 24 of the remaining 26 are classified as "vulnerable" to "critically endangered." Recording the songs of these endangered birds can be a valuable tool for conservation, helping researchers accurately count population sizes and track how birds move in the forest. However, gathering and analyzing these data is both challenging and time-consuming.
To address these challenges, researchers at the University of Hawaiʻi at Hilo, with funding from the Pacific Islands CASC, are combining artificial intelligence (AI) and Native Hawaiian cultural practices. The team has placed acoustic recorders in remote forests on the island of Hawaiʻi to capture a massive amount of soundscape data that is time-intensive to analyze manually. AI can be used to streamline the process, allowing both for faster data analysis and more timely insights for conservation management. Even so, the variability of many Hawaiian bird songs makes it difficult to train AI to automatically detect certain songs. To overcome this, the team has partnered with Google to develop "Perch," an algorithm specifically designed to identify the unique sounds of Hawaiian birds.