Table 7 F1-score (in %) comparison of different fusion strategies.
From: SpectroFusionNet a CNN approach utilizing spectrogram fusion for electric guitar play recognition
Classifier | \(\:{F}_{early,p}\) | \(\:{F}_{late,max}\) | \(\:{F}_{late,wavg}\) | \(\:{F}_{late,concat}\) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
\(\:{S}_{C}+{S}_{G}\) | \(\:{S}_{M}+{S}_{C}\) | \(\:{S}_{M}+{S}_{G}\) | \(\:{S}_{C}+{S}_{G}\) | \(\:{S}_{M}+{S}_{C}\) | \(\:{S}_{M}+{S}_{G}\) | \(\:{S}_{C}+{S}_{G}\) | \(\:{S}_{M}+{S}_{C}\) | \(\:{S}_{M}+{S}_{G}\) | \(\:{S}_{C}+{S}_{G}\) | \(\:{S}_{M}+{S}_{C}\) | \(\:{S}_{M}+{S}_{G}\) | |
Random Forest | 85 | 92 | 90 | 84 | 96 | 96 | 94 | 98 | 99 | 97 | 92 | 99 |
SVM | 74 | 87 | 95 | 85 | 99 | 99 | 97 | 99 | 99 | 96 | 96 | 95 |
KNN | 74 | 83 | 86 | 88 | 96 | 96 | 89 | 94 | 98 | 92 | 98 | 99 |
LMT | 84 | 92 | 96 | 99 | 99 | 99 | 93 | 99 | 99 | 99 | 99 | 99 |
Naive Bayes | 58 | 65 | 67 | 93 | 82 | 80 | 68 | 80 | 80 | 66 | 65 | 75 |
Linear SVM | 88 | 95 | 97 | 99 | 99 | 100 | 96 | 96 | 98 | 99 | 99 | 98 |
MLP | 87 | 90 | 95 | 96 | 97 | 99 | 95 | 98 | 99 | 98 | 99 | 99 |
Decision Tree | 68 | 66 | 59 | 85 | 85 | 86 | 63 | 86 | 85 | 80 | 84 | 84 |
Gradient Boosting | 79 | 88 | 76 | 88 | 89 | 90 | 84 | 96 | 91 | 91 | 93 | 90 |
AdaBoost | 30 | 25 | 26 | 59 | 62 | 72 | 80 | 70 | 86 | 34 | 47 | 45 |
