Table 3 Performance comparison of various audio classification methods on human labeled test set
Model | Accuracy | AUC | Precision | Recall | F-score | Average recall | Top3 accuracy |
|---|---|---|---|---|---|---|---|
GRU | 54.7% | 0.972 | 53.8% | 54.7% | 53.7% | 63.7% | 76.1% |
ResNeXt | 63.8% | 0.984 | 63.4% | 63.8% | 63.3% | 73.1% | 83.4% |
CNNTD-small | 67.4% | 0.9867 | 66.7% | 67.4% | 66.8% | 74.8% | 85.9% |
CNNTD-large | 71.8% | 0.9876 | 70.9% | 71.8% | 71.0% | 77.1% | 88.5% |
PANNs | 73.22% | 0.9546 | 72.75% | 73.22% | 72.88% | 58.31% | 88.41% |