Table 1. Some of the latest and most frequently used datasets in tasks related to bird song classification and detection.
#recs denotes the number of recordings in the dataset; #classes denotes the numbers of classes in each dataset; species tags indicates if there are species specific labels in the recordings stating the presence of specific species in them; annotations denotes the presence of temporal annotations in recordings; duration denotes the approximate duration of each dataset in hours; and other info provides additional information about the characteristics of the dataset.
| Dataset Name | #recs | #classes | species tags | annotations | duration | other info |
|---|---|---|---|---|---|---|
| NIPS4Bplus | 687 | 87 | Yes | Yes | 1 h | |
| freefield1010 | 7,690 | N/A | No | No | 21 h | bird/no-bird tags |
| warblrb10k | 10,000 | N/A | No | No | 28 h | bird/no-bird tags |
| BirdVox-DCASE-20k | 20,000 | N/A | No | No | 55 h | bird/no-bird tags |
| Chernobyl | 6,620 | N/A | No | No | 18 h | bird/no-bird tags |
| PolandNFC | 4,000 | N/A | No | No | 11 h | bird/no-bird tags |
| LifeClef(BirdClef) 2019 | 50,000 | 659 | Yes | No | 350 h | from xeno-canto |
| LifeClef(BirdClef) 2018 | 48,843 | 1500 | Yes | No | 68 h | from xeno-canto |
| BirdVox-full-night | 6 | 25 | No | Yes | 60 h | points in time |