Video Anomaly Detection on XD-Violence
88.53APCPL-VAD
Evaluation Results
| Method | Links | |
|---|---|---|
| CPL-VADFeatures=CLIP2026.02 | 88.53 | |
| HOLMES-VAUSupervision=INSTRU., Explainability=Yes2026.04 | 87.68 | |
| DeSCFeature=CLIP2026.03 | 87.18 | |
| Ex-VAD#Params=-, Expl.=✓, Reas.=✗, Lrn.=-, Obs.=✗2026.07 | 86.52 | |
| MSBTModality=T+R+A+F, Params=11.478M, FLOPs=288.2G2026.02 | 85.92 | |
| π-VAD#Params=-, Expl.=✗, Reas.=✗, Lrn.=-, Obs.=✗2026.07 | 85.37 | |
| BN-WVADFeature=I3D+VGGish2023.11 | 85.26 | |
| Holmes-VADFeature=CLIP2026.03 | 84.96 | |
| Holmes-VADCategory=Prior Explainable VAD Methods2026.05 | 84.96 | |
| Holmes-VADExplainability=true2026.05 | 84.96 | |
| BN-WVADFeature=I3D2023.11 | 84.93 | |
| MSBTModality=R+A+F, Params=3.252M, FLOPs=102.2G2026.02 | 84.78 | |
| VadCLIPMode=Weakly Supervised, Backbone=ViT2025.12 | 84.51 | |
| VadCLIPFeatures=CLIP2026.02 | 84.51 | |
| VadCLIPCategory=Weak2023.08 | 84.51 | |
| VadCLIPFeature=CLIP2026.03 | 84.51 | |
| VADCLIPSupervision=WEAK, Explainability=No2026.04 | 84.51 | |
| VadCLIP#Params=-, Expl.=✗, Reas.=✗, Lrn.=-, Obs.=✗2026.07 | 84.51 | |
| MSBTModality=R+A, Params=0.685M, FLOPs=25.8G2026.02 | 84.12 | |
| STPromptFeature=CLIP2026.03 | 83.97 | |
| Yang et al.Mode=Weakly Supervised, Backbone=ViT2025.12 | 83.68 | |
| TPWNGFeatures=CLIP2026.02 | 83.68 | |
| TPWNGFeature=CLIP2026.03 | 83.68 | |
| SASVenue=arXiv'23, Feature=I3D2023.11 | 83.59 | |
| MACIL-SDVenue=MM'22, Feature=I3D+VGGish2023.11 | 83.4 | |
| MACIL-SD(Full)Modality=R+A, Params=0.678M, FLOPs=56.4G2026.02 | 83.4 | |
| IFS-VADFeatures=CLIP2026.02 | 83.14 | |
| HeadHunt-VADMode=Tuning-Free Multimodal VAD, Backbone=ViT2025.12 | 82.63 | |
| DMUCategory=Weak2023.08 | 82.41 | |
| UR-DMUFeature=CLIP2026.03 | 82.41 | |
| CLIP-TSAMode=Weakly Supervised, Backbone=ViT2025.12 | 82.19 | |
| CLIP-TSAFeatures=CLIP2026.02 | 82.19 | |
| CLIP-TSACategory=Prior Unexplainable VAD Methods2026.05 | 82.19 | |
| CLIP-TSAExplainability=false2026.05 | 82.19 | |
| MACIL-SD(Light)Modality=R+A, Params=0.347M, FLOPs=29.2G2026.02 | 82.17 | |
| CLIP-TSACategory=Weak2023.08 | 82.17 | |
| CLIP-TSAFeature=CLIP2026.03 | 82.17 | |
| HiProbeVADMode=Tuning-Free Multimodal VAD, Backbone=ViT2025.12 | 82.15 | |
| RTFM-BERTFeature=I3D-RGB, Crop=52022.10 | 82.1 | |
| UR-DMUVenue=AAAI'23, Feature=I3D+VGGish2023.11 | 81.77 | |
| Pang et al.Modality=R+A, Params=1.876M, FLOPs=N/A2026.02 | 81.69 | |
| UR-DMUVenue=AAAI'23, Feature=I3D2023.11 | 81.66 | |
| UR-DMUMode=Weakly Supervised, Backbone=I3D2025.12 | 81.66 | |
| UR-DMUFeature=I3D2026.03 | 81.66 | |
| UR-DMU#Params=-, Expl.=✗, Reas.=✗, Lrn.=-, Obs.=✗2026.07 | 81.66 | |
| CU-NetVenue=CVPR'23, Feature=I3D+VGGish2023.11 | 81.43 | |
| MSBTModality=R+F, Params=0.743M, FLOPs=30.8G2026.02 | 80.82 | |
| LA-NetFeature=I3D2026.03 | 80.72 | |
| S3RVenue=ECCV'22, Feature=I3D2023.11 | 80.26 | |
| S3RMode=Weakly Supervised, Backbone=I3D2025.12 | 80.26 | |
| S3RCategory=Prior Unexplainable VAD Methods2026.05 | 80.26 | |
| S3RExplainability=false2026.05 | 80.26 | |
| MGFNFeature=I3D2026.03 | 80.11 | |
| MGFNCategory=Prior Unexplainable VAD Methods2026.05 | 80.11 | |
| MGFNExplainability=false2026.05 | 80.11 | |
| Wu et al.Category=Weak2023.08 | 80 | |
| MACIL-SD(Full)Modality=R+F, Params=0.792M, FLOPs=65.6G, Retrained=true2026.02 | 79.73 | |
| MFGNMode=Weakly Supervised, Backbone=I3D2025.12 | 79.19 | |
| CU-NetVenue=CVPR'23, Feature=I3D2023.11 | 78.74 | |
| Zhang et al.Features=CLIP2026.02 | 78.74 | |
| Wu et al.Modality=R+A, Params=1.539M, FLOPs=N/A2026.02 | 78.66 | |
| [29]Feature=I3D-RGB+audio, Crop=52022.10 | 78.64 | |
| HL-NetVenue=ECCV'20, Feature=I3D+VGGish2023.11 | 78.64 | |
| [14]Feature=VideoSwin-RGB, Crop=52022.10 | 78.59 | |
| MSLNETFeatures=CLIP2026.02 | 78.59 | |
| MSLCategory=Prior Unexplainable VAD Methods2026.05 | 78.58 | |
| MSLExplainability=false2026.05 | 78.58 | |
| RTFMModality=R+A, Params=13.190M, FLOPs=1013.6 G2026.02 | 78.54 | |
| MSLVenue=AAAI'22, Feature=I3D2023.11 | 78.28 | |
| MSLMode=Weakly Supervised, Backbone=I3D2025.12 | 78.28 | |
| MSLFeature=I3D2026.03 | 78.28 | |
| RTFMCategory=Weak2023.08 | 78.27 | |
| RTFMFeature=CLIP2026.03 | 78.27 | |
| AVVDCategory=Weak2023.08 | 78.1 | |
| AVVDFeature=CLIP2026.03 | 78.1 | |
| RTFMFeature=I3D-RGB, Crop=52022.10 | 77.81 | |
| RTFMVenue=ICCV'21, Feature=I3D2023.11 | 77.81 | |
| RTFMMode=Weakly Supervised, Backbone=I3D2025.12 | 77.81 | |
| RTFMFeature=I3D2026.03 | 77.81 | |
| RTFMSupervision=WEAK, Explainability=No2026.04 | 77.81 | |
| RTFMCategory=Prior Unexplainable VAD Methods2026.05 | 77.81 | |
| RTFMExplainability=false2026.05 | 77.81 | |
| Ju et al.Category=Weak2023.08 | 76.57 | |
| COPRABackbone=InternVL2-8B2026.05 | 76.52 | |
| InternVL2-8BMethod Strategy=COPRA2026.05 | 76.52 | |
| [28]Feature=I3D-RGB, Crop=12022.10 | 75.9 | |
| Sultani et al.Category=Weak2023.08 | 75.18 | |
| Sultani et al.Feature=CLIP2026.03 | 75.18 | |
| InternVL2-8BMode=Frozen Baseline2026.05 | 74.32 | |
| InternVL2-8BModel Status=Frozen2026.05 | 74.32 | |
| COPRABackbone=Qwen2-VL-7B-Instruct2026.05 | 73.96 | |
| Qwen2-VL-7B-InstructMethod Strategy=COPRA2026.05 | 73.96 | |
| HL-NetVenue=ECCV'20, Feature=I3D2023.11 | 73.67 | |
| HL-NETSupervision=WEAK, Explainability=No2026.04 | 73.67 | |
| Sultani et al.Venue=CVPR'18, Feature=I3D2023.11 | 73.2 | |
| Wu et al.Mode=Weakly Supervised, Backbone=I3D2025.12 | 73.2 | |
| Sultani et al.Feature=I3D2026.03 | 73.2 | |
| Anom-π#Params=2B, Expl.=✓, Reas.=✓, Lrn.=✓, Obs.=OnD.2026.07 | 72.29 | |
| VERAFeature=CLIP2026.03 | 70.54 | |
| VERA (InternVL2-8B)Category=Prior Explainable VAD Methods2026.05 | 70.54 |