Table 3.
Frame-level AUC performance on all 4 benchmarks using PSNR encoding and SVM classifiers. 80% samples of the test set are used for training SVM. SHT = ShanghaiTech.
| CGAN Stream | Avenue | Ped1 | Ped2 | SHT |
|---|---|---|---|---|
| CGAN-1 | 0.82 | 0.75 | 0.89 | 0.68 |
| CGAN-2 | 0.81 | 0.78 | 0.92 | 0.78 |
| CGAN-3 | 0.79 | 0.64 | 0.75 | 0.73 |
| CGAN-4 | 0.67 | 0.65 | 0.77 | 0.62 |
| CGAN-(1 + 2) | 0.83 | 0.80 | 0.93 | 0.80 |
| CGAN-(3 + 4) | 0.86 | 0.70 | 0.77 | 0.81 |
| CGAN-(1 + 4) | 0.86 | 0.79 | 0.93 | 0.76 |
| CGAN-(1 + 2 + 3) | 0.86 | 0.83 | 0.95 | 0.87 |
| CGAN-(1 + 2 + 4) | 0.83 | 0.83 | 0.93 | 0.83 |
| CGAN-(1 + 2 + 3 + 4) | 0.92 | 0.85 | 0.96 | 0.94 |