Vocalization Classification on MeerKAT 1.0 (5-fold val)
94AP (Focal)animal2vec
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| animal2vecModel architecture=Self-distillation framework, % Training labels=100%2024.06 | 94 | 94 | 92 | 84 | 50 | 80 | 71 | 61 | 50 | 98 | 87 | 80 | 78 | 91 | |
| animal2vecModel architecture=Self-distillation framework, % Training labels=25%2024.06 | 92 | 93 | 91 | 79 | 43 | 73 | 62 | 59 | 39 | 96 | 83 | 77 | 74 | 88 | |
| animal2vecModel architecture=Self-distillation framework, % Training labels=1%2024.06 | 86 | 90 | 88 | 65 | 33 | 57 | 54 | 53 | 41 | 89 | 59 | 74 | 66 | 83 | |
| Transformer baselineModel architecture=Transformer, % Training labels=100%2024.06 | 59 | 49 | 14 | 30 | 7 | 3 | 9 | 9 | 1 | 91 | 12 | 26 | 26 | 30 |