Comparison of action recognition results on PKUMMD dataset with semi-supervised learning approaches.
| Proportion | Models | Architecture | Part I | Part II |
|---|---|---|---|---|
| 1% | Rand-SS | GRU | 31.8 | 10.1 |
| LongT GAN [52] | 35.7 | 12.3 | ||
| MS2L [21] | 36.4 | 13.0 | ||
| EMS2L | 43.0 | 12.5 | ||
| Rand-SS | GCN | 36.8 | 9.0 | |
| EMS2L | 47.1 | 10.0 | ||
| 5% | Rand-SS | GRU | 67.7 | 21.0 |
| EMS2L | 68.8 | 21.2 | ||
| Rand-SS | GCN | 60.4 | 18.2 | |
| EMS2L | 70.3 | 20.0 | ||
| 10% | Rand-SS | GRU | 75.4 | 24.3 |
| LongT GAN [52] | 69.5 | 25.7 | ||
| MS2L [21] | 70.3 | 26.1 | ||
| EMS2L | 76.4 | 28.5 | ||
| Rand-SS | GCN | 76.1 | 26.6 | |
| EMS2L | 76.0 | 23.3 |
| Proportion | Models | Architecture | Part I | Part II |
|---|---|---|---|---|
| 1% | Rand-SS | GRU | 31.8 | 10.1 |
| LongT GAN [ | 35.7 | 12.3 | ||
| MS2L [ | 36.4 | |||
| EMS2L | 12.5 | |||
| Rand-SS | GCN | 36.8 | 9.0 | |
| EMS2L | ||||
| 5% | Rand-SS | GRU | 67.7 | 21.0 |
| EMS2L | ||||
| Rand-SS | GCN | 60.4 | 18.2 | |
| EMS2L | ||||
| 10% | Rand-SS | GRU | 75.4 | 24.3 |
| LongT GAN [ | 69.5 | 25.7 | ||
| MS2L [ | 70.3 | 26.1 | ||
| EMS2L | ||||
| Rand-SS | GCN | |||
| EMS2L | 23.3 |