Activity Recognition on Toyota Smarthome
63.64Mean Per-Class AccuracyAssembleNet++ (object + peer-attention)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AssembleNet++ (object + peer-attention)pre-training=none, modality=object, attention_type=peer-attention2020.08 | 63.64 | 80.64 | |
| AssembleNet++ (object + self-attention)pre-training=none, modality=object, attention_type=self-attention2020.08 | 62.3 | 79.08 | |
| Baseline + self-attentionpre-training=none, attention_type=self-attention2020.08 | 57.84 | 77.59 | |
| Baseline AssembleNet-50pre-training=none2020.08 | 57.42 | 77.77 | |
| I3D + separable STApre-training=Kinetics2020.08 | 54.2 | 75.3 | |
| I3D + NLpre-training=Kinetics2020.08 | 53.6 | — | |
| I3Dpre-training=Kinetics2020.08 | 53.4 | 72 | |
| LSTMpre-training=none2020.08 | 42.5 | — |