FishSoundFinder is an open-source software tool developed by Fisheries and Oceans Canada that uses artificial intelligence to automatically detect fish sounds in underwater acoustic recordings. The system employs Random Forest and Convolutional Neural Network algorithms trained on manually identified fish sounds to identify fish presence in aquatic environments without requiring physical disturbance to the ecosystem.
This system is currently in development and is designed for use by Government of Canada employees working in fisheries research and monitoring. The AI-based approach has been shown to outperform traditional detection methods and has been successfully deployed in British Columbia and other regions including Florida. Because the system does not involve personal information about individuals, it focuses solely on detecting and identifying fish acoustic signals in environmental data.
The tool represents a significant advancement in non-invasive fish population monitoring, allowing scientists to estimate fish presence and activity patterns from passive acoustic recordings rather than through traditional methods that may stress or harm aquatic organisms.
