Audio Classification on AudioSet-20K (test)
37.4mAPATST Base
Evaluation results
| Method | Links | |
|---|---|---|
| ATST BaseProtocol=Fine-tuning, Number of parameters=86M2022.04 | 37.4 | |
| AST-IM + KDPretraining=ImageNet + Knowledge Distillation2021.10 | 34.7 | |
| ATST-ClipProtocol=Linear evaluation2023.06 | 33.8 | |
| ATST-FrameProtocol=Linear evaluation2023.06 | 33 | |
| ATST SmallProtocol=Fine-tuning, Number of parameters=22M2022.04 | 31.5 | |
| SSAST 400Masked Patches=400, Pretraining=Self-Supervised2021.10 | 31 | |
| SSAST-PATCHProtocol=Fine-tuning, Number of parameters=89M2022.04 | 31 | |
| Small-SSASTProtocol=Fine-tuning, Number of parameters=23M2022.04 | 30.8 | |
| SSAST 250Masked Patches=250, Pretraining=Self-Supervised2021.10 | 30.4 | |
| SSAST-FRAMEProtocol=Fine-tuning, Number of parameters=89M2022.04 | 29.2 | |
| AST-AudioSetPretraining=Supervised AudioSet2021.10 | 28.6 | |
| ConformerProtocol=Fine-tuning, Number of parameters=88M2022.04 | 27.6 | |
| AST-ScratchTraining=From scratch2021.10 | 14.8 |