Video Anomaly Detection on ShanghaiTech (test)
0.9814AUCTCA + PEL (Ours)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TCA + PEL (Ours)Supervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.9814 | 0 | — | — | — | — | — | |
| Cho et al.Supervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.976 | — | — | — | — | — | — | |
| CRFDSupervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.9748 | — | — | — | — | — | — | |
| S3RSupervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.9748 | — | — | — | — | — | — | |
| S3RSupervision Protocol=Weakly Supervised, Features=I3D2024.01 | 0.9748 | — | — | — | — | — | — | |
| NL-MILSupervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.9743 | — | — | — | — | — | — | |
| MSLSupervision=Weakly-supervised, Feature=VideoSwin2023.06 | 0.9732 | — | — | — | — | — | — | |
| Yu et al.Supervision Type=Weak Supervision, Features=I3D2022.03 | 0.9727 | — | — | — | — | — | — | |
| RTFMSupervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.9721 | — | — | — | — | — | — | |
| Purwantu et al.Supervision Type=Weak Supervision, Features=TRN2022.03 | 0.9685 | — | — | — | — | — | — | |
| UMLSupervision=Weakly-supervised, Feature=X-CLIP RGB2023.06 | 0.9678 | — | — | — | — | — | — | |
| Our_ws (RTFM)Supervision Protocol=Weakly Supervised, Features=I3D, Note=Degenerated to supervised WS2024.01 | 0.9633 | — | — | — | — | — | — | |
| RTFMSupervision Protocol=Weakly Supervised, Features=I3D, Note=Reproduced results2024.01 | 0.961 | — | — | — | — | — | — | |
| Li et al.Supervision Protocol=Weakly Supervised, Features=I3D2024.01 | 0.9608 | — | — | — | — | — | — | |
| BN-SVPBackbone=C3D2022.03 | 0.96 | — | — | — | — | — | — | |
| MISTSupervision=Weakly-supervised, Feature=I3D RGB2023.06 | 0.9483 | 0.0005 | — | — | — | — | — | |
| MISTSupervision Protocol=Weakly Supervised, Features=I3D2024.01 | 0.9483 | — | — | — | — | — | — | |
| MMILBackbone=C3D2022.03 | 0.9218 | — | — | — | — | — | — | |
| Yu et al.Supervision Type=Weak Supervision, Features=C3D2022.03 | 0.9151 | — | — | — | — | — | — | |
| RTFMSupervision Protocol=Weakly Supervised, Features=C3D, Note=Reproduced results2024.01 | 0.9151 | — | — | — | — | — | — | |
| AR-NetFeature Extractor=I3DConc2021.04 | 0.9124 | 0.1 | — | — | — | — | — | |
| AR-NetSupervision=Weakly-supervised, Feature=RGB+Flow2023.06 | 0.9124 | 0.001 | — | — | — | — | — | |
| Li & VasconcelosBackbone=C3D2022.03 | 0.904 | — | — | — | — | — | — | |
| CLAWSSupervision Type=Weak Supervision, Features=C3D2022.03 | 0.8967 | — | — | — | — | — | — | |
| CLAWSSupervision=Weakly-supervised, Feature=C3D RGB2023.06 | 0.8967 | — | — | — | — | — | — | |
| CLAWSSupervision Protocol=Weakly Supervised, Features=C3D2024.01 | 0.8967 | — | — | — | — | — | — | |
| DAMSupervision Type=Weak Supervision, Features=I3D2022.03 | 0.8822 | — | — | — | — | — | — | |
| Our_wsSupervision Protocol=Unsupervised, Features=I3D, Note=Trained as a whole in framework2024.01 | 0.8818 | — | — | — | — | — | — | |
| Our_wocc (STG-NF)Supervision Protocol=One-Class Classification, Note=Degenerated to supervised OCC2024.01 | 0.8637 | — | — | — | — | — | — | |
| Sultani et al.Backbone=C3D2021.04 | 0.863 | 0.15 | — | — | — | — | — | |
| MIL-RankSupervision=Weakly-supervised, Feature=C3D RGB2023.06 | 0.863 | 0.0015 | — | — | — | — | — | |
| Proposed GCL_WSSupervision Type=Weak Supervision, Features=ResNext2022.03 | 0.8621 | — | — | — | — | — | — | |
| STG-NFVenue=ICCV-23, Image Only=No, Training=Yes2024.07 | 0.859 | — | — | — | — | — | — | |
| STG-NFSupervision Protocol=One-Class Classification2024.01 | 0.859 | — | — | — | — | — | — | |
| Ilse et al.Backbone=C3D2022.03 | 0.8578 | — | — | — | — | — | — | |
| Our_wsSupervision Protocol=Unsupervised, Features=C3D2024.01 | 0.8543 | — | — | — | — | — | — | |
| AR-NetSupervision Protocol=Weakly Supervised, Features=I3D2024.01 | 0.8538 | — | — | — | — | — | — | |
| AnomalyRulerImage Only=Yes, Training=No2024.07 | 0.852 | — | — | — | — | — | — | |
| Object-CentricSetup=Classic VAD2021.08 | 0.849 | — | — | — | — | — | — | |
| SIB-VADType=Pred, Processing Category=Object-centric2025.06 | 0.847 | — | — | — | — | — | — | |
| AnomalyRuler-baseImage Only=Yes, Training=No2024.07 | 0.846 | — | — | — | — | — | — | |
| Sultani et al.Supervision Protocol=Weakly Supervised, Features=I3D, Note=Reproduced results2024.01 | 0.8453 | — | — | — | — | — | — | |
| ADSMType=Likelihood, Processing Category=Object-centric2025.06 | 0.845 | — | — | — | — | — | — | |
| Zhong et al.Backbone=TSN-RGB2021.04 | 0.8444 | — | — | — | — | — | — | |
| GCNSupervision Type=Weak Supervision, Features=TS-RGB2022.03 | 0.8444 | — | — | — | — | — | — | |
| Zhong et al.Backbone=GCN(TSNRGB)2022.03 | 0.8444 | — | — | — | — | — | — | |
| GCNSupervision=Weakly-supervised, Feature=TSN RGB2023.06 | 0.8444 | — | — | — | — | — | — | |
| GCNSupervision Protocol=Weakly Supervised, Features=TSN2024.01 | 0.8444 | — | — | — | — | — | — | |
| SSLVenue=ECCV-22, Image Only=No, Training=Yes2024.07 | 0.843 | — | — | — | — | — | — | |
| JigsawSupervision Protocol=One-Class Classification2024.01 | 0.843 | — | — | — | — | — | — | |
| Noise CleanerSupervision Type=Weak Supervision, Features=C3D2022.03 | 0.8416 | — | — | — | — | — | — | |
| SRFSupervision Type=Weak Supervision, Features=C3D2022.03 | 0.8416 | — | — | — | — | — | — | |
| Zhong et al.Backbone=TSN-Optical-Flow2021.04 | 0.8413 | — | — | — | — | — | — | |
| Zhong et al.Backbone=GCN(TSNOptical Flow)2022.03 | 0.8413 | — | — | — | — | — | — | |
| SSMTL++Supervision Protocol=One-Class Classification2024.01 | 0.838 | — | — | — | — | — | — | |
| AnomalyRuler-baseImage Only=true, Training=false, Source Domain=UB2024.07 | 0.835 | — | — | — | — | — | — | |
| HSCVenue=CVPR-23, Image Only=No, Training=Yes2024.07 | 0.834 | — | — | — | — | — | — | |
| SIB-VADType=Pred, Processing Category=Frame-centric2025.06 | 0.834 | — | — | — | — | — | — | |
| Georgescu et al.Category=Object-centric2021.10 | 0.827 | — | — | — | — | — | — | |
| BAFSetup=Classic VAD2021.08 | 0.827 | — | — | — | — | — | — | |
| BAFVenue=TPAMI-21, Image Only=No, Training=Yes2024.07 | 0.827 | — | — | — | — | — | — | |
| Georgescu et al.Supervision Protocol=One-Class Classification2024.01 | 0.827 | — | — | — | — | — | — | |
| Our_woccSupervision Protocol=Unsupervised, Note=Trained as a whole in framework2024.01 | 0.8257 | — | — | — | — | — | — | |
| Zhang et al.Backbone=C3D2021.04 | 0.825 | 0.1 | — | — | — | — | — | |
| Zhang et al.Supervision Type=Weak Supervision, Features=C3D2022.03 | 0.825 | — | — | — | — | — | — | |
| IBLSupervision=Weakly-supervised, Feature=C3D RGB2023.06 | 0.825 | 0.001 | — | — | — | — | — | |
| Zhang et al.Supervision Protocol=Weakly Supervised, Features=C3D2024.01 | 0.825 | — | — | — | — | — | — | |
| SSMTLSetup=Classic VAD2021.08 | 0.824 | — | — | — | — | — | — | |
| SSMTLSupervision Protocol=One-Class Classification2024.01 | 0.824 | — | — | — | — | — | — | |
| AED-SSMTLType=Multi-tasks, Processing Category=Object-centric2025.06 | 0.824 | — | — | — | — | — | — | |
| Our_woccSupervision Protocol=Unsupervised2024.01 | 0.815 | — | — | — | — | — | — | |
| MoPRL2021.12 | 0.8126 | — | — | — | — | — | — | |
| S3RVenue=ECCV-22, Image Only=No, Training=Yes2024.07 | 0.805 | — | — | — | — | — | — | |
| SSAEType=Pred, Processing Category=Object-centric2025.06 | 0.805 | — | — | — | — | — | — | |
| STG-NF All DataSupervision Protocol=Unsupervised2024.01 | 0.8029 | — | — | — | — | — | — | |
| Proposed GCL_OCCSupervision Type=One Class Classification, Features=ResNext2022.03 | 0.7962 | — | — | — | — | — | — | |
| GCLVenue=CVPR-22, Image Only=No, Training=Yes2024.07 | 0.796 | — | — | — | — | — | — | |
| Wang et al.Category=Miscellaneous2021.10 | 0.793 | — | — | — | — | — | — | |
| CACSetup=Classic VAD2021.08 | 0.793 | — | — | — | — | — | — | |
| CACSupervision Protocol=One-Class Classification2024.01 | 0.793 | — | — | — | — | — | — | |
| MA-PDMType=Pred, Processing Category=Frame-centric2025.06 | 0.792 | — | — | — | — | — | — | |
| FBAEType=Pred, Processing Category=Object-centric2025.06 | 0.792 | — | — | — | — | — | — | |
| AED-MAEType=Patch recon, Processing Category=Frame-centric2025.06 | 0.791 | — | — | — | — | — | — | |
| Proposed GCL_PTSupervision Type=Unsupervised, Features=ResNext2022.03 | 0.7893 | — | — | — | — | — | — | |
| Ours2023.03 | 0.788 | — | — | — | — | — | — | |
| AMP-NetSupervision=Semi-supervised2023.06 | 0.788 | — | — | — | — | — | — | |
| SLMVenue=ICCV-23, Image Only=Yes, Training=Yes2024.07 | 0.788 | — | — | — | — | — | — | |
| MPTType=Multi-tasks, Processing Category=Object-centric2025.06 | 0.788 | — | — | — | — | — | — | |
| Ionescu et al. [12]Category=Object-centric2021.10 | 0.787 | — | — | — | — | — | — | |
| FPDMVenue=ICCV-23, Image Only=Yes, Training=Yes2024.07 | 0.786 | — | — | — | — | — | — | |
| Lu et al. [28]Category=Prediction2021.10 | 0.779 | — | — | — | — | — | — | |
| r-GANSetup=Classic VAD2021.08 | 0.779 | — | — | — | — | — | — | |
| rGANVenue=ECCV-20, Image Only=Yes, Training=Yes2024.07 | 0.779 | — | — | — | — | — | — | |
| rGANVenue=ECCV-20, Image Only=true, Training=true, Source Domain=ShanghaiTech2024.07 | 0.779 | — | — | — | — | — | — | |
| CT-D2GANSetup=Classic VAD2021.08 | 0.777 | — | — | — | — | — | — | |
| CT-D2GANSupervision Protocol=One-Class Classification2024.01 | 0.777 | — | — | — | — | — | — | |
| VADMambaType=Hybrid, Processing Category=Frame-centric2025.06 | 0.77 | — | — | — | — | — | — | |
| RUVADSupervision Type=One Class Classification2022.03 | 0.7667 | — | — | — | — | — | — | |
| Multipath-Pred.Setup=Classic VAD2021.08 | 0.766 | — | — | — | — | — | — | |
| Zhong et al.Backbone=C3D2021.04 | 0.7644 | — | — | — | — | — | — |