Video Anomaly Detection on ShanghaiTech standard (test)
97.98Frame-Level AUCProDisc-VAD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ProDisc-VADReference=This work, Feature=ViT-B/162025.05 | 97.98 | — | — | |
| Cho et al.Reference=CVPR23, Feature=I3D2025.05 | 97.6 | — | — | |
| DARReference=TIFS22, Feature=I3D2025.05 | 97.54 | — | — | |
| VadCLIPReference=AAAI24, Feature=ViT-B/162025.05 | 97.49 | — | — | |
| S3RReference=ECCV22, Feature=I3D2025.05 | 97.48 | — | — | |
| RTFMReference=ICCV21, Feature=I3D2025.05 | 97.21 | — | — | |
| MSLReference=AAAI22, Feature=I3D2025.05 | 96.08 | — | — | |
| MISTReference=CVPR21, Feature=I3D2025.05 | 94.83 | — | — | |
| RTFMReference=ICCV21, Feature=C3D2025.05 | 91.51 | — | — | |
| CLAWSReference=ECCV20, Feature=C3D2025.05 | 89.67 | — | — | |
| STG-NF2022.11 | 85.9 | 52.1 | 82.4 | |
| Sultani et al.Reference=CVPR18, Feature=I3D2025.05 | 85.33 | — | — | |
| AnomalyRulerZero-shot=false2025.03 | 85.2 | — | — | |
| object-centric convolutional auto-encoders2018.12 | 84.9 | — | — | |
| Wang et al.Input=Pre-trained object detection results2022.11 | 84.2 | 22.4 | 60.8 | |
| Barbalau et al. (SSMTL++v2)Input=Pre-trained object detection results2022.11 | 83.8 | 47.1 | 85.6 | |
| Georgescu et al. + SSPCABInput=Pre-trained object detection results2022.11 | 83.6 | 40.6 | 83.5 | |
| MULDEZero-shot=false2025.03 | 81.3 | — | — | |
| AnyAnomalyZero-shot=true, Context-aware=true2025.03 | 79.7 | — | — | |
| AnyAnomalyCross-domain training=false, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 79.7 | — | — | |
| Zaheer et al.2022.11 | 79.6 | — | — | |
| MA-PDMZero-shot=false2025.03 | 79.2 | — | — | |
| AED-MAEZero-shot=false2025.03 | 79.1 | — | — | |
| SLMZero-shot=false2025.03 | 78.8 | — | — | |
| Shibao et al.Cross-domain training=true, Few-shot adaptation=false, Auxiliary datasets=true2025.03 | 78.7 | — | — | |
| FPDMZero-shot=false2025.03 | 78.6 | — | — | |
| rGANCross-domain training=true, Few-shot adaptation=true, Auxiliary datasets=false2025.03 | 77.9 | — | — | |
| AnyAnomalyZero-shot=true, Context-aware=false2025.03 | 77.2 | — | — | |
| Sultani et al.pre-trained model=true2018.12 | 76.5 | — | — | |
| Zhong et al.Reference=CVPR19, Feature=C3D2025.05 | 76.44 | — | — | |
| AccI-VADZero-shot=false2025.03 | 76.2 | — | — | |
| Markovitz et al.Input=Pose data2022.11 | 76.1 | — | — | |
| Rodrigues et al.Input=Pose data2022.11 | 76 | — | — | |
| DLAN-ACZero-shot=false2025.03 | 74.7 | — | — | |
| MPNZero-shot=false2025.03 | 73.8 | — | — | |
| USTN-DSCZero-shot=false2025.03 | 73.8 | — | — | |
| MPNCross-domain training=true, Few-shot adaptation=true, Auxiliary datasets=false2025.03 | 73.8 | — | — | |
| AMMC-NetZero-shot=false2025.03 | 73.7 | — | — | |
| STEAL-NetZero-shot=false2025.03 | 73.7 | — | — | |
| Morais et al.Input=Pose data2022.11 | 73.4 | — | — | |
| Liu et al. [21]2018.12 | 72.8 | — | — | |
| zxVADCross-domain training=true, Few-shot adaptation=false, Auxiliary datasets=true2025.03 | 71.6 | — | — | |
| Video-ChatGPTCross-domain training=false, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 69.1 | — | — | |
| Luo et al.2018.12 | 68 | — | — | |
| ZS ImageBindCross-domain training=false, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 61.3 | — | — | |
| ZS CLIPCross-domain training=false, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 60.9 | — | — | |
| Hasan et al.2018.12 | 60.9 | — | — | |
| LLaVA-1.5Cross-domain training=false, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 59.6 | — | — | |
| JigsawCross-domain training=true, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 59.3 | — | — | |
| STEAL-NetCross-domain training=true, Few-shot adaptation=false, Auxiliary datasets=false2025.03 | 51.7 | — | — |