Research Article
A New Method of Pedestrian Abnormal Behavior Detection Based on Attention Guidance
Table 4
Comparative experiments of the attention-directed abnormal behavior detection model (EER%).
| Method | Model | Ped1 | Ped2 | Avenue |
| — | Unmasking [28] | 31.2 | — | — | Level set method [29] | 25.8 | 14.0 | — | Two-stream R-ConvVAE [30] | 32.4 | 15.5 | 27.5 | Two stream with OFF [24] | — | — | 22.9 |
| Reconstruction | CAE [31] | 43.1 | 24.5 | 27.0 | AE_mem_recon [13] | 37.6 | 18.7 | 27.4 | Att_AE_recon | 37.0 | 17.2 | 24.6 |
| Prediction | SRNN [32] | — | — | 24.7 | GAN_pred [33] | 24.0 | 15.6 | 22.5 | AE_mem_pred [13] | 26.9 | 15.0 | 23.3 | ST-3DCAE [34] | 25.1 | 21.8 | 24.9 | sRNN-AE [35] | - | 14.9 | 23.2 | Two stream 3DAE [9] | 27.2 | 15.5 | 24.8 | Memory_att [17] | — | — | 21.6 | PST [36] | — | — | 20.8 | Att_AE_pred | 23.9 | 13.7 | 19.5 |
|
|