A Santa Cruz, California based company is using big data and deep learning to improve conservation efforts. California birdwatchers can go a lifetime without seeing the globally endangered marbled murelett bird and with now with remote acoustic sensors and deep learning, biologists are now able to analyze the audio of the bird, and keep better track the populations of species that were previously hard to monitor.
In a recent article, the CEO of Conservation Metrics mentioned, “we need to improve conservation by improving wildlife monitoring. Counting plants and animals is really tricky business.” And they are doing so with the help of TITAN GPUs.
The company created spectrograms of bird calls compiled from acoustic data they gathered and is converted into a visual representation — time runs along the x-axis, audio frequency along the y-axis — and then an algorithm searches for patterns to isolate specific bird calls. In the soundscape, many different species are making noise at once, but the computer has identified a marbled murrelet call within the clip. It’s the upside-down horseshoe pattern running along the lower third: