Temporal Action Detection on THUMOS-14 (test)
80.9mAP@tIoU=0.5AdaTAD
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| AdaTADBackbone=VideoMAEv2-g, E2E=true, Flow=false, Mem=50.6, Resolution=1536x224^2, Note=Last half of backbone layers adapted2023.11 | 80.9 | — | — | 89.7 | 86.7 | 71 | 56.1 | 76.9 | — | |
| AdaTAD†Backbone=VideoMAEv2-g, E2E=true, Flow=false, Mem=43.6, Resolution=1536x224^22023.11 | 79.4 | — | — | 89.6 | 85.9 | 67.6 | 53.8 | 75.4 | — | |
| AdaTADBackbone=VideoMAEv2-g, E2E=true, Flow=false, Mem=29.92023.11 | 78.9 | — | — | 89.5 | 85.8 | 67.3 | 52.6 | 74.8 | — | |
| AdaTADBackbone=VideoMAE-H, E2E=true, Flow=false, Mem=19.22023.11 | 78.6 | — | — | 88.9 | 85.3 | 66.9 | 52.5 | 74.4 | — | |
| AdaTAD*backbone=VMAE-L, reproduced_by_authors=true2026.04 | 77.6 | — | — | 87.5 | 83.7 | 66.5 | 51.9 | 73.5 | — | |
| AdaTAD* + MOSSbackbone=VMAE-L2026.04 | 77.6 | — | — | 88.1 | 84.5 | 67.7 | 52.8 | 74.1 | — | |
| AdaTADBackbone=VideoMAE-L, E2E=true, Flow=false, Mem=11.02023.11 | 76.7 | — | — | 87.7 | 84.1 | 66.4 | 52.4 | 73.5 | — | |
| AdaTADbackbone=VMAE-L2026.04 | 76.7 | — | — | 87.7 | 84.1 | 66.4 | 52.4 | 73.5 | — | |
| ADI-DiffFeature=I3D2024.04 | 76.5 | — | — | 84.9 | 81.5 | 63 | 48 | 70.8 | — | |
| AdaTAD* + MOSSbackbone=VMAE-B2026.04 | 75.7 | — | — | 87.3 | 82.9 | 64.8 | 50.3 | 72.2 | — | |
| AdaTADBackbone=VideoMAE-B, E2E=true, Flow=false, Mem=4.92023.11 | 75.3 | — | — | 87 | 82.4 | 63.8 | 49.2 | 71.5 | — | |
| AdaTADbackbone=VMAE-B2026.04 | 75.3 | — | — | 87 | 82.4 | 63.8 | 49.2 | 71.5 | — | |
| AdaTAD*backbone=VMAE-B, reproduced_by_authors=true2026.04 | 75.3 | — | — | 85.7 | 82.3 | 63.4 | 50.1 | 71.4 | — | |
| TriDetFeature=I3D2024.04 | 72.9 | — | — | 83.6 | 80.1 | 62.4 | 47.4 | 69.3 | — | |
| TriDetBackbone=I3D, E2E=false, Flow=true2023.11 | 72.9 | — | — | 83.6 | 80.1 | 62.4 | 47.4 | 69.3 | — | |
| TriDetbackbone=I3D2026.04 | 72.9 | — | — | 83.6 | 80.1 | 62.4 | 47.4 | 69.3 | — | |
| DyFADetLatency CPU(s)=0.49, Latency GPU(ms)=58.7, Params (M)=18.82024.07 | 72.7 | — | — | — | — | — | — | 69.2 | — | |
| ASLBackbone=I3D, E2E=false, Flow=true2023.11 | 71.7 | — | — | 83.1 | 79 | 59.7 | 45.8 | 67.9 | — | |
| DyFADet*Latency CPU(s)=0.16, Latency GPU(ms)=44.7, Params (M)=12.72024.07 | 71.7 | — | — | — | — | — | — | 67.8 | — | |
| AdaTADBackbone=VideoMAE-S, E2E=true, Flow=false, Mem=2.52023.11 | 71.6 | — | — | 84.5 | 80.2 | 60.9 | 46.9 | 68.8 | — | |
| DiffTADFeature=I3D2024.04 | 71.2 | — | — | 74.9 | 72.8 | 62.9 | 58.5 | 68 | — | |
| ActionFormerFeature=I3D2024.04 | 71 | — | — | 82.1 | 77.8 | 59.4 | 43.9 | 66.8 | — | |
| ActionFormerBackbone=I3D, E2E=false, Flow=true2023.11 | 71 | — | — | 82.1 | 77.8 | 59.4 | 43.9 | 66.8 | — | |
| ActionFormerLatency CPU(s)=0.57, Latency GPU(ms)=68.4, Params (M)=29.22024.07 | 71 | — | — | — | — | — | — | 66.8 | — | |
| ActionFormerbackbone=I3D2026.04 | 71 | — | — | 82.1 | 77.8 | 59.4 | 43.9 | 66.8 | — | |
| AdaTADBackbone=SlowFast-R50, E2E=true, Flow=false, Mem=4.32023.11 | 69.4 | — | — | 81 | 76.2 | 59 | 44.5 | 66 | — | |
| ActionFormerBackbone=SlowFast-R50, E2E=false, Flow=true2023.11 | 65.2 | — | — | 78.7 | 73.3 | 54.6 | 39.7 | 62.3 | — | |
| PlusTADType=One-stage, Backbone=R50-SlowOnly, RGB-Only=true, FLOPs=533.1G, Resolution=160, Temporal window=6,192, Mechanism=Anchor-free2022.05 | 63.5 | — | — | 75.5 | 70.8 | 50.9 | 37.4 | 59.6 | — | |
| BasicTADBackbone=SlowOnly-R50, E2E=true, Flow=false, Mem=122023.11 | 63.5 | — | — | 75.5 | 70.8 | 50.9 | 37.4 | 59.6 | — | |
| TALLFormerBackbone=VideoSwin-B, E2E=true, Flow=false, Mem=292023.11 | 63.2 | — | — | 76 | — | — | 34.5 | 59.2 | — | |
| TALLFormerbackbone=VSwin-B2026.04 | 63.2 | — | — | 76 | — | — | 34.5 | 59.2 | — | |
| Re2TALBackbone=Re2 VideoSwin-T, E2E=true, Flow=false, Mem=242023.11 | 62.4 | — | — | 77 | 71.5 | 49.7 | 36.3 | 59.4 | — | |
| PlusTADType=One-stage, Backbone=R50-SlowOnly, RGB-Only=true, FLOPs=519.3G, Resolution=160, Temporal window=6,192, Mechanism=Anchor-based2022.05 | 62 | — | — | 72.3 | 68.4 | 52.4 | 37 | 58.4 | — | |
| Self-DETRFeature=I3D2024.04 | 60 | — | — | 74.6 | 69.5 | 47.6 | 31.8 | 56.7 | — | |
| PlusTADType=One-stage, Backbone=R50-SlowOnly, RGB-Only=true, FLOPs=136.4G, Resolution=112, Temporal window=3,96, Mechanism=Anchor-based2022.05 | 58.6 | — | — | 68.4 | 65 | 49.2 | 33.5 | 54.9 | — | |
| PlusTADType=One-stage, Backbone=R50-SlowOnly, RGB-Only=true, FLOPs=151.5G, Resolution=112, Temporal window=3,96, Mechanism=Anchor-free2022.05 | 57.6 | — | — | 70.4 | 65.5 | 46 | 33.2 | 54.5 | — | |
| ReActType=One-stage, FLOPs (G)=0.682022.07 | 57.1 | — | — | 69.2 | 65 | 47.8 | 35.6 | 55 | — | |
| ReActFeature=3DCNN2024.04 | 57.1 | — | — | 69.2 | 65 | 47.8 | 35.6 | 55 | — | |
| TAGSType=Proposal Free, Backbone=I3D2022.07 | 57 | — | — | 68.6 | 63.8 | 46.3 | 31.8 | 52.8 | — | |
| TAGSFeature=I3D2024.04 | 57 | — | — | 68.6 | 63.8 | 46.3 | 31.8 | 52.8 | — | |
| MUSESType=Anchor, Backbone=I3D2022.07 | 56.9 | — | — | 68.9 | 64 | 46.3 | 31 | 53.4 | — | |
| MUSESFeature=I3D2024.04 | 56.9 | — | — | 68.9 | 64 | 46.3 | 31 | — | — | |
| STPTFeature=STPT2024.04 | 56.4 | — | — | 70.6 | 65.7 | 44.6 | 30.5 | 53.6 | — | |
| TadTRType=Multi-stage, Backbone=R50-SlowFast, RGB-Only=true, FLOPs=475.0G2022.05 | 56 | — | — | 69.4 | 64.3 | 46.4 | 34.9 | 54.2 | — | |
| E2E-TADBackbone=SlowFast-R50, E2E=true, Flow=false, Mem=122023.11 | 56 | — | — | 69.4 | 64.3 | 46.4 | 34.9 | 54.2 | — | |
| SP-TADType=Multi-stage, Backbone=I3D, RGB-Only=false2022.05 | 55.9 | — | — | 69.2 | 63.3 | 45.7 | 33.4 | 53.5 | — | |
| AFSDType=Two-Stage, Flow=true2021.07 | 55.5 | — | — | 67.3 | 62.4 | 43.7 | 31.1 | — | — | |
| AFSDType=One-stage, FLOPs (G)=5.12022.07 | 55.5 | — | — | 67.3 | 62.4 | 43.7 | 31.1 | 52 | — | |
| AFSDType=Multi-stage, Backbone=I3D, RGB-Only=false, FLOPs=2780.0G2022.05 | 55.5 | — | — | 67.3 | 62.4 | 43.7 | 31.1 | 52 | — | |
| AFSDFeature=I3D2024.04 | 55.5 | — | — | 67.3 | 62.4 | 43.7 | 31.1 | 52 | — | |
| AFSDBackbone=I3D, E2E=true, Flow=true, Mem=122023.11 | 55.5 | — | — | 67.3 | 62.4 | 43.7 | 31.1 | 52 | — | |
| DaoTADType=One-stage, Backbone=R50-SlowOnly, RGB-Only=true, FLOPs=133.3G2022.05 | 54.4 | — | — | 63.2 | 59.7 | 45.6 | 32.2 | 51 | — | |
| ContextLocType=Two-stage, FLOPs (G)=3.12022.07 | 54.3 | — | — | 68.3 | 63.8 | 41.8 | 26.2 | 50.9 | — | |
| ContextLocType=Actioness, Backbone=I3D2022.07 | 54.3 | — | — | 68.3 | 63.8 | 41.8 | 26.2 | — | — | |
| ContextLocFeature=I3D2024.04 | 54.3 | — | — | 68.3 | 63.8 | 41.8 | 26.2 | 50.9 | — | |
| DCANVideo-level classifier=UntrimmedNet2021.12 | 54.1 | — | — | 68.2 | 62.7 | 43.9 | 32.6 | — | — | |
| DCANType=Multi-stage, Backbone=TSN, RGB-Only=false2022.05 | 54.1 | — | — | 68.2 | 62.7 | 43.9 | 32.6 | 52.3 | — | |
| DaoTADType=One-Stage, Flow=false2021.07 | 53.8 | — | — | 62.8 | 59.5 | 43.6 | 30.1 | — | — | |
| DaoTADType=One-stage, Backbone=R50-I3D, RGB-Only=true, FLOPs=206.7G2022.05 | 53.8 | — | — | 62.8 | 59.5 | 43.6 | 30.1 | 50 | — | |
| P-GCNFeature=R(2+1)D2024.04 | 53.5 | — | — | 69.1 | 63.3 | 40.5 | 26 | 50.5 | — | |
| MLTPNType=One-Stage, Flow=true2021.07 | 53.3 | — | — | 66 | 62.6 | 37 | 21.1 | — | — | |
| VSGNType=Two-stage2022.07 | 52.4 | — | — | 66.7 | 60.4 | 41 | 30.4 | 50.2 | — | |
| VSGNType=Anchor, Backbone=I3D2022.07 | 52.4 | — | — | 66.7 | 60.4 | 41 | 30.4 | 50.1 | — | |
| VSGNFeature=TSN2024.04 | 52.4 | — | — | 66.7 | 60.4 | 41 | 30.4 | 50.2 | — | |
| C-TCNType=One-Stage, Flow=true2021.07 | 52.1 | — | — | 68 | 62.3 | — | — | — | — | |
| AEI-BClassifier=P-GCN2021.10 | 52.1 | — | — | 67.3 | 60.6 | 37.8 | 22.4 | — | — | |
| AEI-GClassifier=P-GCN2021.10 | 52 | — | — | 67.6 | 60.4 | 37.9 | 22.3 | — | — | |
| RTD-NetType=Two-stage2022.07 | 51.9 | — | — | 68.3 | 62.3 | 38.8 | 23.7 | 49 | — | |
| RTD-NetType=Actioness, Backbone=I3D2022.07 | 51.9 | — | — | 68.3 | 62.3 | 38.8 | 23.7 | — | — | |
| RTD-NetFeature=I3D2024.04 | 51.9 | — | — | 68.3 | 62.3 | 38.8 | 23.7 | 49 | — | |
| G-TADType=One-Stage, Flow=true2021.07 | 51.6 | — | — | 66.4 | 60.4 | 37.6 | 22.9 | — | — | |
| GTADClassifier=P-GCN2021.10 | 51.6 | — | — | 66.4 | 60.4 | 37.6 | 22.9 | — | — | |
| GTAD+PGCNType=Mixed, Backbone=I3D2022.07 | 51.6 | — | — | 66.4 | 60.4 | 37.6 | 22.9 | 47.8 | — | |
| GTADClassifier=P-GCN [66], Year=CVPR20, Feature=2Stream2022.10 | 51.6 | — | — | 66.4 | 60.4 | 37.6 | 22.9 | 47.78 | — | |
| G-TADType=Multi-stage, Backbone=TSN, RGB-Only=false2022.05 | 51.6 | — | — | 66.4 | 60.4 | 37.6 | 22.9 | 47.8 | — | |
| PBRNetType=One-Stage, Flow=true2021.07 | 51.3 | — | — | 58.5 | 54.6 | 41.8 | 29.5 | — | — | |
| PBR-NetType=Anchor, Backbone=I3D2022.07 | 51.3 | — | — | 58.5 | 54.6 | 41.8 | 29.5 | — | — | |
| PBRNetType=One-stage, Backbone=I3D, RGB-Only=false2022.05 | 51.3 | — | — | 58.5 | 54.6 | 41.8 | 29.5 | 47.1 | — | |
| PBRNetFeature=I3D2024.04 | 51.3 | — | — | 58.5 | 54.6 | 41.8 | 29.5 | — | — | |
| AOE-NetClassifier=P-GCN [66], Feature=C3D2022.10 | 50.9 | — | — | 67.1 | 60.6 | 37.4 | 23.5 | 47.89 | — | |
| MRClassifier=P-GCN [66], Year=ECCV20, Feature=2Stream2022.10 | 50.1 | — | — | 66.29 | 60.99 | — | — | — | — | |
| CSA+BMNType=Two-stage2022.07 | 49.2 | — | — | 64.4 | 58 | 38.2 | 27.8 | 47.5 | — | |
| TadTrType=One-stage, FLOPs (G)=0.752022.07 | 49.2 | — | — | 62.4 | 57.4 | 37.8 | 26.3 | 46.6 | — | |
| BMN-CSAFeature=TSN2024.04 | 49.2 | — | — | 64.4 | 58 | 38.2 | 27.8 | 47.7 | — | |
| TadTRFeature=I3D2024.04 | 49.2 | — | — | 62.4 | 57.4 | 37.8 | 26.3 | 46.6 | — | |
| TadTRBackbone=I3D, E2E=false, Flow=true2023.11 | 49.2 | — | — | 62.4 | 57.4 | 37.8 | 26.3 | 46.6 | — | |
| GCNsType=Two-Stage, Flow=true2021.07 | 49.1 | — | — | 63.6 | 57.8 | — | — | — | — | |
| BSNClassifier=P-GCN2021.10 | 49.1 | — | — | 63.6 | 57.8 | — | — | — | — | |
| P-GCNType=Two-stage, FLOPs (G)=4.42022.07 | 49.1 | — | — | 63.6 | 57.8 | — | — | — | — | |
| BSNClassifier=P-GCN [66], Year=ECCV18, Feature=I3D2022.10 | 49.1 | — | — | 63.6 | 57.8 | — | — | — | — | |
| AOE-NetClassifier=UntrimmedNet [68], Feature=C3D2022.10 | 48.4 | — | — | 63.4 | 57.3 | 38.8 | 25.8 | 46.74 | — | |
| TSA-NetFeature=P3D2024.04 | 46.9 | — | — | 61.2 | 55.9 | 36.1 | 25.2 | 45.1 | — | |
| TAGSType=Proposal Free, Backbone=TS2022.07 | 46.5 | — | — | 61.4 | 52.9 | 38.1 | 27 | 44 | — | |
| PointTAD2022.10 | 46.2 | — | — | 62.6 | 55.9 | 35.3 | 22.8 | 44.6 | — | |
| ABNClassifier=Unet2022.03 | 46.12 | — | — | 59.87 | 53.95 | 37.04 | 25.56 | — | — | |
| ABNClassifier=UntrimmedNet [68], Year=IEEE-Access 21, Feature=C3D2022.10 | 46.1 | — | — | 59.9 | 54 | 37 | 25.6 | 44.51 | — | |
| A2NetType=One-Stage, Flow=true2021.07 | 45.5 | — | — | 58.6 | 54.1 | 32.5 | 17.2 | — | — | |
| A2NetType=One-stage, FLOPs (G)=30.42022.07 | 45.5 | — | — | 58.6 | 54.1 | 32.5 | 17.2 | 41.6 | — | |
| A2NetType=Mixed, Backbone=I3D2022.07 | 45.5 | — | — | 58.6 | 54.1 | 32.5 | 17.2 | 41.6 | — | |
| A2NetType=One-stage, Backbone=I3D, RGB-Only=false2022.05 | 45.5 | — | — | 58.6 | 54.1 | 32.5 | 17.2 | 41.6 | — |