基于多尺度外观运动融合的视频异常检测
|
|
赵明华,吕雨萱,吕佳豪,陈逸飞,石程,胡静
|
Video anomaly detection based on multi-scale appearance and motion fusion
|
|
Minghua ZHAO,Yuxuan LYU,Jiahao LYU,Yifei CHEN,Cheng SHI,Jing HU
|
|
| 表 2 所提方法与先进方法在AUC指标上的对比 |
| Tab.2 Comparison of proposed method and advanced methods in terms of AUC metric |
|
| 类型 | 方法 | AUC/% | | Ped2 | Avenue | SHT | | 重建 | MemAE[1] | 94.1 | 83.3 | 71.2 | | Astrid等[36] | 94.8 | 84.9 | 72.5 | | GMFC-VAE[37] | 92.2 | 83.4 | — | | CR-AE[38] | 95.6 | — | 73.1 | | MESDnet[35] | 95.6 | 86.3 | 73.2 | | 预测 | Park等[9] | 97.0 | 88.5 | 70.5 | | Liu等[8] | 95.4 | 85.1 | 72.8 | | MPN[19] | 96.9 | 89.5 | 73.8 | | Conv-VRNN[12] | 96.1 | 85.8 | — | | FSCN[13] | 92.8 | 85.5 | — | | AMMC-Net[11] | 96.6 | 86.6 | 73.7 | | Li等[16] | 96.5 | 87.1 | 73.6 | | PDM-Net[39] | 97.7 | 88.1 | 74.2 | | GroupGAN[40] | 96.6 | 88.1 | 74.2 | | AMAE[41] | 97.4 | 88.2 | 73.6 | | 本研究 | 99.1 | 88.5 | 74.5 |
|
|
|