Audio Classification on FSD50K (train/test)
58.27mAPprotobin
Evaluation Results
| Method | Links | |
|---|---|---|
| protobinBackbone=BEATs, Input Category=Token Map (Prot), Pooling Method=protobin2025.09 | 58.27 | |
| protobinBackbone=Dasheng, Input Category=Token Map (Prot), Pooling Method=protobin2025.09 | 57.31 | |
| protobinBackbone=SSLAM, Input Category=Token Map (Prot), Pooling Method=protobin2025.09 | 56.99 | |
| protobinBackbone=EAT, Input Category=Token Map (Prot), Pooling Method=protobin2025.09 | 55.64 | |
| protobinBackbone=A-MAE, Input Category=Token Map (Prot), Pooling Method=protobin2025.09 | 49.69 | |
| protoBackbone=A-MAE, Input Category=Token Map (Prot), Pooling Method=proto2025.09 | 49.65 | |
| protobinBackbone=ASiT, Input Category=Token Map (Prot), Pooling Method=protobin2025.09 | 46.7 | |
| mhcaBackbone=A-MAE, Input Category=Token Map (Att.), Pooling Method=mhca2025.09 | 45.17 | |
| linearBackbone=A-MAE, Input Category=[cls] Baseline, Pooling Method=linear2025.09 | 19.71 |