Research Article
Abnormal Detection in Big Data Video with an Improved Autoencoder
Table 1
Comparison with the state-of-the-art methods in terms of AUC.
| | Method | Avenue | ShanghaiTech |
| | MPPCA + SF | 56.2% | — | | MDT | 77.4% | — | | Conv-AE | 80.0% | 60.9% | | Conv3D-AE | 80.9% | — | | Stacked RNN | 81.7% | 68.0% | | ConvLSTM-AE | 77.0% | — | | MemNormality | 88.5% | 70.5% | | ClusterAE | 86.0% | 73.3% | | AbnormalGAN | — | 72.4% | | Pred + Recon | 85.1% | 73.0% | | The proposed method | 85.9% | 75.4% |
|
|
