The team of researchers under the Bird Audio Detection Challenge program managed to teach artificial intelligence to recognize birds by their singing.
The new function will help correctly calculate the halo of living of birds of one or another species in different parts of the world. This is important for tracking the trend towards the disappearance of birds and taking measures to preserve populations in certain regions, which is important, given the scope of deforestation, agriculture and the trends of climate change.
In the process of work, experts involved 30 programs based on machine training, which were practiced on the records of birds singing from the alienation zone of the Chernobyl nuclear power plant. A wide variety of birds is concentrated in this huge quiet territory, and their sounds are not distorted and do not mix with the rumble of human activity – the voices of people, the city, the sounds of cars, factories and other things.
During tests, artificial intelligence showed recognition accuracy in 89%. So far, AI is not able to surpass the possibilities of organitologists who can distinguish birds by singing, but the neural network can be used in any corners of the planet, and it is not against the rain or work around the clock. But it is likely that after some time, AI will be able to boast of the best result than that of experts. If only the birds continued to exist in the natural habitat, and not in cells.
Source: Sciencemag.Org
The team of researchers under the Bird Audio Detection Challenge program managed to teach artificial intelligence to recognize birds by their singing.
The new function will help correctly calculate the halo of living of birds of one or another species in different parts of the world. This is important for tracking the trend towards the disappearance of birds and taking measures to preserve populations in certain regions, which is important, given the scope of deforestation, agriculture and the trends of climate change.
In the process of work, experts involved 30 programs based on machine training, which were practiced on the records of birds singing from the alienation zone of the Chernobyl nuclear power plant. A wide variety of birds is concentrated in this huge quiet territory, and their sounds are not distorted and do not mix with the rumble of human activity – the voices of people, the city, the sounds of cars, factories and other things.
During tests, artificial intelligence showed recognition accuracy in 89%. So far, AI is not able to surpass the possibilities of organitologists who can distinguish birds by singing, but the neural network can be used in any corners of the planet, and it is not against the rain or work around the clock. But it is likely that after some time, AI will be able to boast of the best result than that of experts. If only the birds continued to exist in the natural habitat, and not in cells.
Source: Sciencemag.Org