Temporal Action Detection on ActivityNet 1.3
64.2mAP@0.5MambaTAD
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MambaTADE2E=true, Mem=16.7G, Backbone=VideoMAEv2-G2025.11 | 64.2 | 45 | 11.3 | 43.8 | |
| ESTF-SSMBackbone=VideoMAEv2, E2E=true, Mem=28.6G2026.04 | 63.8 | 44.6 | 11.2 | 43.9 | |
| MambaTADE2E=true, Mem=9.3G, Backbone=VideoMAE-H2025.11 | 63.3 | 44.4 | 11.2 | 43.1 | |
| MambaTADE2E=false, Backbone=IV6B2025.11 | 63.1 | 44.2 | 11 | 42.8 | |
| MambaTADBackbone=InternVideo2, E2E=true, Mem=30.7G2026.04 | 63.1 | 44.2 | 11 | 43.5 | |
| ActionMambaTraining Type=Head only, Head Type=Anchor-free, Feature=InternVideo22025.05 | 62.4 | 43.5 | 10.2 | 42 | |
| VideoMambaSuiteE2E=false, Backbone=IV6B2025.11 | 62.4 | 43.5 | 10.2 | 42 | |
| ActionMambaBackbone=InternVideo2, E2E=false2026.04 | 62.2 | 37.1 | 9.6 | 41.7 | |
| VideoMambaBackbone=InternVideo2, E2E=false2026.04 | 62.1 | 43.2 | 9.5 | 41.6 | |
| DIGITTraining Type=Head only, Head Type=Query-based, Feature=InternVideo22025.05 | 62 | 43.1 | 11.3 | 42 | |
| AdaTADTraining Type=Full, Feature=VideoMAEv2-g2025.05 | 61.7 | 43.4 | 10.9 | 41.9 | |
| ActionFormerTraining Type=Head only, Head Type=Anchor-free, Feature=InternVideo22025.05 | 61.5 | 44.6 | 12.7 | 41.2 | |
| InternVideo2*E2E=false, Backbone=IV6B2025.11 | 61.5 | 44.6 | 12.7 | 41.2 | |
| TE-TADTraining Type=Head only, Head Type=Query-based, Feature=InternVideo22025.05 | 61.3 | 41.8 | 10.9 | 41.1 | |
| AdaTAD++Backbone=VideoMAEv2, E2E=true, Mem=30.9G2026.04 | 60.3 | 43.2 | 10.1 | 41 | |
| TadTRTraining Type=Head only, Head Type=Query-based, Feature=InternVideo2, MDGE=true, CAID=true2025.05 | 60.2 | 41 | 11.2 | 40.5 | |
| MambaTADE2E=false, Backbone=R(2+1)D2025.11 | 60.2 | 41.3 | 9.7 | 40.2 | |
| AdaTADBackbone=VideoMAEv2, E2E=true, Mem=29.9G2026.04 | 59.8 | 42 | 9.7 | 40.8 | |
| BDRC-NetBackbone=InternVideo2, E2E=false2026.04 | 59.6 | 38.9 | 9.8 | 41.9 | |
| LFAFE2E=false, Backbone=I3D†2025.11 | 59.3 | 39.8 | 7.8 | 39 | |
| DIGITTraining Type=Head only, Head Type=Query-based, Feature=InternVideo2, NMS=false2025.05 | 58.9 | 43.4 | 11.4 | 41.3 | |
| CLTDR-GMGBackbone=InternVideo2, E2E=false2026.04 | 58.3 | 40.2 | 9.4 | 39.5 | |
| DyFADetBackbone=TSP+IV12026.04 | 58.19 | 39.3 | 8.63 | 38.62 | |
| Zhu et al.Backbone=I3D2024.03 | 58.1 | 36.3 | 6.2 | 35.2 | |
| DyFADetTraining Type=Head only, Head Type=Anchor-free, Feature=I3D / R(2+1)D2025.05 | 58.1 | 39.6 | 8.4 | 38.5 | |
| DyFADetE2E=false, Backbone=R(2+1)D2025.11 | 58.1 | 39.6 | 8.4 | 38.5 | |
| AdaTADE2E=true, Mem=19.2G, Backbone=VideoMAE-H2025.11 | 58 | 40.6 | 9.8 | 39.4 | |
| AdaTADE2E=true, Mem=11.0G, Backbone=VideoMAE-L2025.11 | 57.7 | 40.6 | 10.1 | 39.2 | |
| AdaTADType=ANN, Backbone=ViT-L [17]2026.06 | 57.69 | 40.56 | 10.13 | 39.22 | |
| TadTRTraining Type=Head only, Head Type=Query-based, Feature=InternVideo22025.05 | 57.1 | 38.8 | 11 | 38.2 | |
| AdaTADE2E=true, Mem=4.9G, Backbone=VideoMAE-B2025.11 | 56.8 | 39.4 | 9.7 | 38.4 | |
| AdaTADType=ANN, Backbone=ViT-B [17]2026.06 | 56.77 | 39.35 | 9.71 | 38.39 | |
| TriDetBackbone=SlowFast, Flow=false2023.12 | 56.7 | 39.3 | 11.7 | 38.6 | |
| TAGSBackbone=I3D2024.03 | 56.3 | 36.8 | 9.6 | 36.5 | |
| AdaTADE2E=true, Mem=2.5G, Backbone=VideoMAE-S2025.11 | 56.2 | 39 | 9.1 | 37.9 | |
| AdaTADType=ANN, Backbone=ViT-S [17]2026.06 | 56.15 | 38.99 | 9.07 | 37.85 | |
| DiffTADE2E=false, Backbone=I3D†2025.11 | 56.1 | 36.9 | 9 | 36.1 | |
| DiffTAD (10-step)Backbone=I3D2026.04 | 56.1 | 36.9 | 9 | 36.1 | |
| Baseline(ANN)Type=ANN, Backbone=ViT-S [17]2026.06 | 56.09 | 38.71 | 8.86 | 37.64 | |
| SpikeTADType=ANN2SNN, Backbone=ViT-S [17], T (time-steps in the backbone)=162026.06 | 56.04 | 38.97 | 8.75 | 37.42 | |
| ContextLocBackbone=I3D2024.03 | 56 | 35.2 | 3.6 | 34.2 | |
| ViT-TADE2E=true, Backbone=ViT-B2025.11 | 55.9 | 38.5 | 8.8 | 37.4 | |
| ViT-TADBackbone=ViT-B, Flow=false, Resolution=160x1602023.12 | 55.87 | 38.47 | 8.8 | 37.4 | |
| ViT-TADType=ANN, Backbone=ViT-B [17]2026.06 | 55.87 | 38.47 | 8.8 | 37.4 | |
| SpikeTADType=ANN2SNN, Backbone=ViT-S [17], T (time-steps in the backbone)=82026.06 | 55.87 | 37.99 | 8.69 | 37.05 | |
| SpikeTAD(ANN)Type=ANN, Backbone=ViT-S [17]2026.06 | 55.83 | 37.92 | 8.67 | 37.12 | |
| ViT-TADTraining Type=Full, Feature=ViT-B2025.05 | 55.8 | 38.5 | 8.8 | 37.4 | |
| CausalTADBackbone=TSP2026.04 | 55.62 | 38.51 | 9.4 | 37.46 | |
| ADSFormerE2E=false, Backbone=VM22025.11 | 55.3 | 38.4 | 8.4 | 37.1 | |
| Re2TALType=ANN, Backbone=Re2SlowFast-101 [13]2026.06 | 55.3 | 37.9 | 9.1 | 37 | |
| SpikeTADType=ANN2SNN, Backbone=ViT-S [17], T (time-steps in the backbone)=42026.06 | 55.26 | 37.21 | 8.8 | 36.53 | |
| Re2TALBackbone=SlowFast-R101, Flow=false, Resolution=224x2242023.12 | 55.25 | 37.86 | 9.05 | 37.01 | |
| LiquidTADBackbone=TSP2026.04 | 55.18 | 38.1 | 8.43 | 37.06 | |
| ViT-TADBackbone=ViT-S, Flow=false, Resolution=160x1602023.12 | 55.09 | 37.81 | 8.75 | 36.69 | |
| ViT-TADType=ANN, Backbone=ViT-S [17]2026.06 | 55.09 | 37.81 | 8.75 | 36.69 | |
| ActionFormerBackbone=TSP2026.04 | 55.08 | 38.27 | 8.91 | 37.07 | |
| TriDetBackbone=TSP2026.04 | 54.89 | 38.2 | 8.21 | 36.96 | |
| Re2TALE2E=true, Mem=24G, Backbone=Re2VideoSwin-T2025.11 | 54.8 | 37.8 | 9 | 36.8 | |
| BSFTALBackbone=InternVideo2, E2E=false2026.04 | 54.8 | 37.5 | 9.1 | 37.1 | |
| VSGNBackbone=TSP2026.04 | 54.8 | 37.35 | 9.8 | 36.89 | |
| Re2TALBackbone=Swin-T, Flow=false, Resolution=224x2242023.12 | 54.75 | 37.81 | 9.03 | 36.8 | |
| Re2TALType=ANN, Backbone=Re2VideoSwin-T [13]2026.06 | 54.75 | 37.81 | 9.03 | 36.8 | |
| TriDetTraining Type=Head only, Head Type=Anchor-free, Feature=I3D / R(2+1)D2025.05 | 54.7 | 38 | 8.4 | 36.8 | |
| ActionFormerTraining Type=Head only, Head Type=Anchor-free, Feature=I3D / R(2+1)D2025.05 | 54.7 | 37.8 | 8.4 | 36.6 | |
| ActionFormerE2E=false, Backbone=R(2+1)D2025.11 | 54.7 | 37.8 | 8.4 | 36.6 | |
| TriDetE2E=false, Backbone=SlowFast2025.11 | 54.7 | 38 | 8.4 | 36.8 | |
| TridetType=ANN, Backbone=I3D [15]2026.06 | 54.7 | 38 | 8.4 | 36.8 | |
| TriDetBackbone=I3D, Flow=true2023.12 | 54.5 | 36.8 | 11.5 | 36.8 | |
| DIGITTraining Type=Head only, Head Type=Query-based, Feature=I3D / R(2+1)D2025.05 | 54.4 | 38.2 | 10.7 | 37.3 | |
| VSGN (w/o GCN)Backbone=TSP2026.04 | 54.4 | 36.88 | 9.15 | 36.27 | |
| DIGITTraining Type=Head only, Head Type=Query-based, Feature=I3D / R(2+1)D, NMS=false2025.05 | 54.3 | 38.4 | 10.6 | 37.2 | |
| TCANetE2E=false, Backbone=SlowFast2025.11 | 54.3 | 39.1 | 8.4 | 37.6 | |
| ActionFormerType=ANN, Backbone=SlowFast-R50 [43]2026.06 | 54.26 | 37.04 | 8.13 | 36.02 | |
| TE-TADTraining Type=Head only, Head Type=Query-based, Feature=I3D / R(2+1)D2025.05 | 54.2 | 38.1 | 10.6 | 37.1 | |
| TALLFormerBackbone=Swin-B, Flow=false, Resolution=224x2242023.12 | 54.1 | 36.2 | 7.9 | 35.6 | |
| TALLFormerTraining Type=Full, Feature=Swin-B2025.05 | 54.1 | 36.2 | 7.9 | 35.6 | |
| TALLFormerE2E=true, Mem=29G, Backbone=Swin2025.11 | 54.1 | 36.2 | 7.9 | 35.6 | |
| TALLFormerType=ANN, Backbone=VideoSwin-B [45]2026.06 | 54.1 | 36.2 | 7.9 | 35.6 | |
| PBRNetBackbone=I3D, Flow=true, Resolution=96x962023.12 | 53.96 | 34.97 | 8.98 | 35.01 | |
| TadTR*Feature=TSP, Ensemble=true, focal loss=true, GFLOPs=0.0382021.06 | 53.62 | 37.52 | 10.56 | 36.75 | |
| TadTRBackbone=TSP2026.04 | 53.62 | 37.52 | 10.56 | 36.75 | |
| TadTRTraining Type=Head only, Head Type=Query-based, Feature=I3D / R(2+1)D2025.05 | 53.6 | 37.5 | 10.6 | 36.8 | |
| TadTRE2E=false, Backbone=TSN†2025.11 | 53.6 | 37.5 | 10.6 | 36.8 | |
| ActionFormerBackbone=I3D2024.03 | 53.5 | 36.2 | 8.2 | 35.6 | |
| ActionFormerBackbone=I3D, Flow=true2023.12 | 53.5 | 36.2 | 8.2 | 35.6 | |
| ActionFormerType=ANN, Backbone=I3D [15]2026.06 | 53.5 | 36.2 | 8.2 | 35.6 | |
| BMNBackbone=TSP2026.04 | 52.9 | 37.3 | 9.67 | 36.4 | |
| TadTR*Feature=I3D, Ensemble=true, focal loss=true, GFLOPs=0.0382021.06 | 52.83 | 37.05 | 10.83 | 36.11 | |
| DualDETRBackbone=I3D2024.03 | 52.6 | 35 | 7.8 | 34.3 | |
| DualDETRTraining Type=Head only, Head Type=Query-based, Feature=I3D2025.05 | 52.6 | 35 | 7.8 | 34.3 | |
| TSIBackbone=TSP2026.04 | 52.44 | 35.57 | 9.8 | 35.36 | |
| BMN-CSABackbone=TSN2024.03 | 52.4 | 36.2 | 5.2 | 35.4 | |
| VSGNBackbone=TSN2024.03 | 52.4 | 36 | 8.4 | 35.1 | |
| AFSDBackbone=I3D2024.03 | 52.4 | 35.3 | 6.5 | 34.4 | |
| AFSDBackbone=I3D, Flow=true, Resolution=96x962023.12 | 52.4 | 35.3 | 6.5 | 34.4 | |
| AFSDTraining Type=Full, Feature=I3D2025.05 | 52.4 | 35.3 | 6.5 | 34.4 | |
| VSGNE2E=false, Backbone=TSN†2025.11 | 52.4 | 36 | 8.4 | 35.1 | |
| AFSDE2E=true, Mem=12G, Backbone=I3D†2025.11 | 52.4 | 35.3 | 6.5 | 34.4 | |
| AFSDFeature=I3D, Ensemble=true, GFLOPs=3.32021.06 | 52.38 | 35.27 | 6.47 | 34.39 | |
| VSGNBackbone=TSN, Flow=true2023.12 | 52.38 | 36.01 | 8.37 | 35.07 |