Table 3 Comparison of classification accuracy (%) and weak pose reconstruction error with different numbers of Gaussian processes and different vocabulary size
From: Human action recognition based on estimated weak poses
Number ofGPs | ||||||||
|---|---|---|---|---|---|---|---|---|
3 | 6 | 10 | 15 | 20 | 25 | 30 | ||
Voc size | 5 | 73.9 | 86.8 | 86.3 | 86.0 | 86.1 | 86.1 | 85.6 |
10 | 67.7 | 83.6 | 82.9 | 83.0 | 84.4 | 84.2 | 84.2 | |
15 | 64.1 | 83.9 | 82.6 | 80.8 | 85.4 | 83.9 | 83.7 | |
20 | 64.7 | 79.0 | 77.5 | 79.7 | 78.4 | 84.2 | 82.2 | |
Mean error | 0.399 | 0.304 | 0.241 | 0.200 | 0.169 | 0.146 | 0.127 | |