If you live in a city, it may not be difficult to identify a bird solely by its song, since the feathered species that have adapted to live in our metropolis are not that numerous. In nature it’s a different story and, although recognizing birds by their songs can be seen as an idle hobby, it’s an important ability if you are an ornithologist surveying a forest to identify an endangered bird in a chorus of hundreds of songs. With that problem in mind a team of coders has developed a tool that is also fun to use.
The Bird Sounds experiment, powered by Google Creative lab, is a visualization of thousands of bird songs organized by their similarities. An AI decided which songs were similar and filed them in a grid in which you can “glide” with your cursor and hear a swarm of birdcalls; you can also look for a specific bird to listen to.
The envisioned future application for this technology is to walk with a recording device and being able to distinguish and count not only birds, but any class of animal sound to help scientists survey populations and species in different ecosystems.
Take a look and have fun discovering which birds you can identify.
Source: AI Experiments