Segmented audio recordings of birdsong

100 minutes of audio recordings of 10 bird species singing + json files with markup. The audio recordings themselves are cut and divided into the appropriate folders in order to avoid confusion.
The use of the dataset for commercial purposes is allowed only with the consent of the copyright holder (manager@labelme.ru )

All audio recordings are segmented and checked for the absence of abnormal noises and sounds. The layout specialists adhered to strict terms of reference:

1) Any abnormal sounds should not get into the markup: the singing of other birds, the noise of trees, people talking, and any other sounds different than the pure singing of a bird of this species.
2) If several birds of the same species can be heard singing at the same time on the audio recording, it is not necessary to include such segments in the markup, since it is required to mark the pure singing of one bird of each biological species.
3) It is desirable that the duration of the segments should not be less than half a second, but at the same time the extra sounds described earlier should not fall into the segment.

The dataset can be used to create a neural network that generates nature sounds, or a bird Shazam that will determine which bird sings in real time.