Target Sound Extraction on ESC-50 (test)
12.37SISDRiCLAPSep-hybrid
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CLAPSep-hybridModality=Audio, Shots=10, Polarity=P+N2024.02 | 12.37 | 13.29 | |
| CLAPSep-hybridModality=Audio, Shots=5, Polarity=P+N2024.02 | 12.34 | 13.26 | |
| CLAPSep-hybridModality=Audio+Text, Shots=10, Polarity=P+N2024.02 | 12.24 | 13.21 | |
| CLAPSep-hybridModality=Audio, Shots=1, Polarity=P+N2024.02 | 11.94 | 12.89 | |
| CLAPSep-hybridModality=Audio, Shots=10, Polarity=N2024.02 | 11.75 | 12.64 | |
| CLAPSep-hybridModality=Audio, Shots=10, Polarity=P2024.02 | 11.73 | 12.79 | |
| CLAPSep-hybridModality=Audio, Shots=5, Polarity=P2024.02 | 11.65 | 12.72 | |
| CLAPSep-hybridModality=Audio, Shots=5, Polarity=N2024.02 | 11.61 | 12.49 | |
| CLAPSep-hybridModality=Audio+Text, Shots=10, Polarity=P2024.02 | 11.3 | 12.46 | |
| CLAPSep-hybridModality=Audio+Text, Shots=10, Polarity=N2024.02 | 11.29 | 12.28 | |
| CLAPSep-textModality=Audio+Text, Shots=10, Polarity=P+N2024.02 | 11.17 | 12.12 | |
| CLAPSep-hybridModality=Audio, Shots=1, Polarity=P2024.02 | 10.88 | 12.08 | |
| CLAPSep-hybridModality=Audio, Shots=1, Polarity=N2024.02 | 10.8 | 11.76 | |
| USS-ResUNet30Modality=Audio, Shots=10, Polarity=P2024.02 | 8.68 | 9.51 | |
| SuDORM-RF++Multiplier=1.0x, Params (1e6)=2.72, GFLOPs=16.23, Mem. (GB)=0.99, Time (sec)=0.272021.03 | 8.6 | — | |
| SuDORM-RF++Scale=1.0x, GFLOPs=2.11, Memory (GB)=0.79, Inference Time (sec)=0.172021.03 | 8.6 | — | |
| SuDORM-RF++ GCMultiplier=1.0x, Params (1e6)=0.3, GFLOPs=2.72, Mem. (GB)=1.21, Time (sec)=0.352021.03 | 8.5 | — | |
| SuDORM-RF++Scale=1.0x, Global Convolution=true, GFLOPs=0.69, Memory (GB)=0.87, Inference Time (sec)=0.22021.03 | 8.5 | — | |
| SuDORM-RFMultiplier=1.0x, Params (1e6)=2.72, GFLOPs=17.19, Mem. (GB)=0.99, Time (sec)=0.272021.03 | 8.4 | — | |
| SuDORM-RFScale=1.0x, GFLOPs=2.45, Memory (GB)=0.79, Inference Time (sec)=0.172021.03 | 8.4 | — | |
| USS-ResUNet30Modality=Audio, Shots=5, Polarity=P2024.02 | 8.38 | 9.29 | |
| SuDoRM-RFScale=1.0x, Parameters (millions)=2.66, GFLOPs (Inference)=2.52, GFLOPs (Backward update)=2.56, Memory (GB) (Inference)=0.61, Memory (GB) (Backward update)=0.86, Time (sec) (Inference)=0.67, Time (sec) (Backward update)=0.382020.07 | 8.35 | — | |
| Two-Step TDCNParameters (millions)=8.63, GFLOPs (Inference)=7.09, GFLOPs (Backward update)=7.23, Memory (GB) (Inference)=0.99, Memory (GB) (Backward update)=1.17, Time (sec) (Inference)=1.05, Time (sec) (Backward update)=0.302020.07 | 8.22 | — | |
| Two-Step TDCNParams (1e6)=8.63, GFLOPs=28.36, Mem. (GB)=1.17, Time (sec)=0.222021.03 | 8.2 | — | |
| Two-Step TDCNGFLOPs=7.11, Memory (GB)=1.03, Inference Time (sec)=0.242021.03 | 8.2 | — | |
| SuDoRM-RFScale=0.5x, Parameters (millions)=1.42, GFLOPs (Inference)=1.54, GFLOPs (Backward update)=1.56, Memory (GB) (Inference)=0.40, Memory (GB) (Backward update)=0.45, Time (sec) (Inference)=0.36, Time (sec) (Backward update)=0.212020.07 | 8.12 | — | |
| SuDORM-RFMultiplier=0.5x, Params (1e6)=1.42, GFLOPs=9.47, Mem. (GB)=0.33, Time (sec)=0.112021.03 | 8.1 | — | |
| SuDORM-RFScale=0.5x, GFLOPs=1.51, Memory (GB)=0.49, Inference Time (sec)=0.132021.03 | 8.1 | — | |
| SuDoRM-RFScale=0.25x, Parameters (millions)=0.79, GFLOPs (Inference)=1.06, GFLOPs (Backward update)=1.07, Memory (GB) (Inference)=0.30, Memory (GB) (Backward update)=0.25, Time (sec) (Inference)=0.29, Time (sec) (Backward update)=0.132020.07 | 7.93 | — | |
| SuDORM-RFMultiplier=0.25x, Params (1e6)=0.79, GFLOPs=5.62, Mem. (GB)=0.28, Time (sec)=0.082021.03 | 7.9 | — | |
| SuDORM-RFScale=0.25x, GFLOPs=1.04, Memory (GB)=0.3, Inference Time (sec)=0.092021.03 | 7.9 | — | |
| ConvTasNetParameters (millions)=5.05, GFLOPs (Inference)=5.23, GFLOPs (Backward update)=5.30, Memory (GB) (Inference)=0.65, Memory (GB) (Backward update)=0.88, Time (sec) (Inference)=0.90, Time (sec) (Backward update)=0.332020.07 | 7.74 | — | |
| ConvTasNetParams (1e6)=5.05, GFLOPs=21.7, Mem. (GB)=0.68, Time (sec)=0.142021.03 | 7.7 | — | |
| ConvTasNetGFLOPs=5.16, Memory (GB)=0.61, Inference Time (sec)=0.152021.03 | 7.7 | — | |
| DemucsParameters (millions)=415.09, GFLOPs (Inference)=3.43, GFLOPs (Backward update)=10.34, Memory (GB) (Inference)=2.24, Memory (GB) (Backward update)=8.77, Time (sec) (Inference)=0.53, Time (sec) (Backward update)=0.362020.07 | 7.23 | — | |
| DPRNNParameters (millions)=2.63, GFLOPs (Inference)=48.89, GFLOPs (Backward update)=48.90, Memory (GB) (Inference)=2.23, Memory (GB) (Backward update)=3.40, Time (sec) (Inference)=3.98, Time (sec) (Backward update)=0.602020.07 | 7.2 | — | |
| DemucsParams (1e6)=415.09, GFLOPs=17.65, Mem. (GB)=8.77, Time (sec)=0.242021.03 | 7.2 | — | |
| DPRNNParams (1e6)=2.63, GFLOPs=135.88, Mem. (GB)=3.28, Time (sec)=0.322021.03 | 7.2 | — | |
| DemucsGFLOPs=3.42, Memory (GB)=1.95, Inference Time (sec)=0.582021.03 | 7.2 | — | |
| DPRNNGFLOPs=48.81, Memory (GB)=2.27, Inference Time (sec)=2.182021.03 | 7.2 | — | |
| USS-ResUNet30Modality=Audio, Shots=1, Polarity=P2024.02 | 7.08 | 8.17 | |
| CLAPSep-textModality=Audio, Shots=10, Polarity=P+N2024.02 | 6.16 | 6.93 | |
| C-SuDORM-RF++Multiplier=0.5x, Params (1e6)=2.81, GFLOPs=11.43, Mem. (GB)=0.29, Time (sec)=0.092021.03 | 4.6 | — | |
| C-SuDORM-RF++Scale=0.5x, Causal=true, GFLOPs=2.14, Memory (GB)=0.39, Inference Time (sec)=0.082021.03 | 4.6 | — | |
| C-SuDORM-RF++Multiplier=0.25x, Params (1e6)=1.63, GFLOPs=6.27, Mem. (GB)=0.17, Time (sec)=0.052021.03 | 4.5 | — | |
| C-SuDORM-RF++Scale=0.25x, Causal=true, GFLOPs=1.25, Memory (GB)=0.3, Inference Time (sec)=0.052021.03 | 4.5 | — |