Temporal Action Localization on ActivityNet 1.3
38.5Average mAPConTrans
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ConTransMode=RGB+Flow, Backbone=I3D2026.05 | 38.5 | 60.2 | 42.4 | 22.3 | |
| SMILEBackbone=ViT-B, Pre-trained=K4002025.04 | 38 | — | — | — | |
| MDVLM-TAL-8BStage=1-stage, Category=Generative2026.05 | 37.9 | 58.2 | 38.8 | 9.8 | |
| SIGMABackbone=ViT-B, Pre-trained=K4002025.04 | 37.7 | — | — | — | |
| MGMBackbone=ViT-B, Pre-trained=K4002025.04 | 37.6 | — | — | — | |
| VideoMAEBackbone=ViT-B, Pre-trained=K4002025.04 | 37.3 | — | — | — | |
| MGMAEBackbone=ViT-B, Pre-trained=K4002025.04 | 37.3 | — | — | — | |
| CLTDR-GMGStage=1-stage, Category=Discriminative2026.05 | 37.1 | 55 | 38 | 8.6 | |
| ActionFormer + Re²TALBackbone=Re²Slowfast-101, End-to-End=true, Optical Flow=false, Memory (GB)=6.82022.11 | 37.01 | 55.25 | 37.86 | 9.05 | |
| ActionFormer + Re²TALBackbone=Re²VSwin-Tiny, End-to-End=true, Optical Flow=false, Memory (GB)=2.42022.11 | 36.8 | 54.75 | 37.81 | 9.03 | |
| TriDetStage=1-stage, Category=Discriminative2026.05 | 36.8 | 54.7 | 38 | 8.4 | |
| TransGMCStage=1-stage, Category=Discriminative2026.05 | 36.7 | 54.8 | 37.6 | 8.5 | |
| ActionFormerStage=1-stage, Category=Discriminative2026.05 | 36.6 | 54.7 | 37.8 | 8.4 | |
| ConTransMode=RGB, Backbone=CLIP2026.05 | 36.6 | 54.5 | 41.8 | 15.5 | |
| TAGSStage=1-stage, Category=Discriminative2026.05 | 36.5 | 56.3 | 36.8 | 9.6 | |
| STALEMode=RGB+Flow, Backbone=I3D2026.05 | 36.4 | 56.5 | 36.7 | 9.5 | |
| DiffTAD (10-step)Stage=1-stage, Category=Generative2026.05 | 36.1 | 56.1 | 36.9 | 9 | |
| ActionFormerBackbone=Slowfast-101, End-to-End=false, Optical Flow=false2022.11 | 36.09 | 53.98 | 37 | 8.87 | |
| TFFormerStage=1-stage, Category=Discriminative2026.05 | 35.8 | 54.4 | 36.7 | 7.5 | |
| ActionFormerBackbone=I3D, End-to-End=false, Optical Flow=true2022.11 | 35.6 | 53.5 | 36.2 | 8.2 | |
| TaLLFormerBackbone=VSwin-Base, End-to-End=true, Optical Flow=false, Memory (GB)=292022.11 | 35.6 | 54.1 | 36.2 | 7.9 | |
| TALLFormerStage=1-stage, Category=Discriminative2026.05 | 35.6 | 54.1 | 36.2 | 7.9 | |
| TCANetStage=2-stage, Category=Discriminative2026.05 | 35.5 | 52.3 | 36.7 | 6.9 | |
| ActionFormerBackbone=VSwin-Tiny, End-to-End=false, Optical Flow=false2022.11 | 35.17 | 53.83 | 35.82 | 7.27 | |
| VSGNBackbone=TSN, End-to-End=false, Optical Flow=true2022.11 | 35.07 | 52.38 | 36.01 | 8.37 | |
| PBRNetBackbone=I3D, End-to-End=true, Optical Flow=false2022.11 | 35.01 | 53.96 | 34.97 | 8.98 | |
| BSN++Supervision=Full2021.04 | 34.9 | 51.3 | 35.7 | 8.3 | |
| AFSDBackbone=I3D, End-to-End=true, Optical Flow=true, Memory (GB)=122022.11 | 34.4 | 52.4 | 35.3 | 6.5 | |
| STALEMode=RGB, Backbone=CLIP2026.05 | 34.3 | 54.3 | 34 | 7.7 | |
| BC-GNNBackbone=TSN, End-to-End=false, Optical Flow=true2022.11 | 34.26 | 50.56 | 34.75 | 9.37 | |
| ContextLocStage=2-stage, Category=Discriminative2026.05 | 34.2 | 56 | 35.2 | 3.6 | |
| TSIBackbone=TSN, End-to-End=false, Optical Flow=true2022.11 | 34.15 | 51.18 | 35.02 | 6.59 | |
| G-TADSupervision=Full2021.04 | 34.1 | 50.4 | 34.6 | 9 | |
| G-TADStage=2-stage, Category=Discriminative2026.05 | 34.1 | 50.4 | 34.6 | 9 | |
| G-TADBackbone=TSN, End-to-End=false, Optical Flow=true2022.11 | 34.09 | 50.36 | 34.6 | 9.02 | |
| MUSESMode=RGB+Flow, Backbone=I3D2026.05 | 34 | 50 | 34.9 | 6.5 | |
| BMNSupervision=Full2021.04 | 33.9 | 50.1 | 34.8 | 8.3 | |
| BMNStage=2-stage, Category=Discriminative2026.05 | 33.9 | 50.1 | 34.8 | 8.3 | |
| BMNBackbone=TSN, End-to-End=false, Optical Flow=true2022.11 | 33.85 | 50.07 | 34.78 | 8.29 | |
| B-IIMode=RGB, Backbone=CLIP2026.05 | 32.7 | 51.5 | 33.3 | 6.6 | |
| ReActStage=1-stage, Category=Discriminative2026.05 | 32.6 | 49.6 | 33 | 8.6 | |
| DiffTAD baseline (10-step)Stage=2-stage, Category=Generative2026.05 | 32.4 | 51 | 32.9 | 9 | |
| P-GCNSupervision=Full2021.04 | 31.1 | 48.3 | 33.2 | 3.3 | |
| RTD-NetStage=2-stage, Category=Discriminative2026.05 | 30.8 | 47.2 | 30.7 | 8.6 | |
| B-IIIMode=RGB+Flow, Backbone=I3D2026.05 | 30.8 | 47.2 | 30.7 | 8.6 | |
| BSNSupervision=Full2021.04 | 30 | 46.5 | 30 | 8 | |
| EffPromptMode=RGB, Backbone=CLIP2026.05 | 27.3 | 44 | 27 | 5.1 | |
| HR-Prosupervision=Point-level (Weak)2023.08 | 27.1 | 42.8 | 27.2 | 8 | |
| LACPsupervision=Point-level (Weak)2023.08 | 25.1 | 40.4 | 24.6 | 5.7 | |
| A2NetMode=RGB, Backbone=I3D2026.05 | 24.8 | 39.6 | 25.7 | 2.8 | |
| ACM-NetSupervision=Weak2021.04 | 24.6 | 40.1 | 24.2 | 6.2 | |
| SSNSupervision=Full2021.04 | 24 | 39.1 | 23.5 | 5.5 | |
| CRRC-Netsupervision=Point-level (Weak)2023.08 | 24 | 39.8 | 24.1 | 5.9 | |
| ACSNetSupervision=Weak2021.04 | 23.9 | 36.3 | 24.2 | 5.8 | |
| A2CL-PTSupervision=Weak2021.04 | 22.5 | 36.8 | 22.5 | 5.2 | |
| BasNetSupervision=Weak2021.04 | 22.2 | 34.5 | 22.5 | 4.9 | |
| TAL-NetBackbone=I3D, End-to-End=false, Optical Flow=true2022.11 | 20.22 | 38.23 | 18.3 | 1.3 | |
| TAL-NetStage=2-stage, Category=Discriminative2026.05 | 20.2 | 38.2 | 18.3 | 1.3 | |
| TALNetMode=RGB+Flow, Backbone=I3D2026.05 | 20.2 | 38.2 | 18.3 | 1.3 | |
| B-IMode=RGB, Backbone=CLIP2026.05 | 18.2 | 28.2 | 18.3 | 3.7 | |
| MAANSupervision=Weak2021.04 | — | 33.7 | 21.9 | 5.5 | |
| R-C3DBackbone=C3D, End-to-End=true, Optical Flow=false2022.11 | — | 26.8 | — | — | |
| STPNSupervision=Weak2021.04 | — | 29.3 | 16.9 | 2.6 | |
| TSMSupervision=Weak2021.04 | — | 30.3 | 19 | 4.5 |