Research Article
Anomaly Detection in Moving Crowds through Spatiotemporal Autoencoding and Additional Attention
Table 3
Performance of the different approaches across datasets (frame-level AUC).
| Source dataset | Avenue | Ped1 | Ped2 |
| Target dataset | Ped1 | Ped2 | Avenue | Ped2 | Avenue | Ped1 |
| Conv-AE | 19.8% | 15.6% | 17.8% | 67.4% | 17.5% | 68.8% |
| ConvLSTM-AE | 23.6% | 19.2% | 18.3% | 81.2% | 19.5% | 78.5% |
| Our approach | 80.3% | 84.4% | 68.2% | 85.3% | 69.6% | 81.6% |
|
|