Research Article
Indian Classical Dance Action Identification and Classification with Convolutional Neural Networks
Table 4
Recognition rates of offline dance data identification with different classifiers.
| Classifier | Average recognition rates (%) of offline dance data | Batch I training | Batch II training | Batch III training | Testing with the same dataset | Testing with different dataset | Testing with the same dataset | Testing with different dataset | Testing with the same dataset | Testing with different dataset |
| ANN [15] | 83.42 | 80.01 | 85.03 | 81.61 | 86.49 | 82.11 | Deep ANN [16] | 87.39 | 83.62 | 88.01 | 84.92 | 89.49 | 86.99 | SVM [13] | 75.93 | 71.31 | 78.35 | 74.48 | 80.46 | 76.78 | Adaboost [14] | 80.19 | 76.49 | 80.98 | 77.29 | 81.76 | 79.09 | AGM [17] | 87.20 | 83.16 | 87.89 | 84.63 | 88.23 | 85.05 | LeNet [25] | 88.14 | 85.49 | 88.55 | 86.32 | 87.92 | 86.85 | VGG [27] | 89.98 | 87.02 | 90.12 | 88.41 | 88.76 | 88.02 | Our proposed CNN architecture | 92.14 | 90.88 | 92.83 | 91.17 | 93.33 | 92.47 |
|
|