Temporal Action Detection on THUMOS '14
87.7mAP@0.3AdaTAD
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| AdaTADType=ANN, Backbone=ViT-L [17]2026.06 | 87.7 | 84.1 | 76.7 | 66.4 | 52.4 | 73.5 | — | — | |
| AdaTADType=ANN, Backbone=ViT-B [17]2026.06 | 87 | 82.4 | 75.3 | 63.8 | 49.2 | 71.5 | — | — | |
| ViT-TADType=ANN, Backbone=ViT-B [17]2026.06 | 85.1 | 80.9 | 74.2 | 61.8 | 45.4 | 69.5 | — | — | |
| AdaTADType=ANN, Backbone=ViT-S [17]2026.06 | 84.5 | 80.2 | 71.6 | 60.9 | 46.9 | 68.8 | — | — | |
| TriDetBackbone=I3D2024.03 | 83.6 | 80.1 | 72.9 | 62.4 | 47.4 | 69.3 | — | — | |
| TridetType=ANN, Backbone=I3D [15]2026.06 | 83.6 | 80.1 | 72.9 | 62.4 | 47.4 | 69.3 | — | — | |
| DualDETRBackbone=I3D2024.03 | 82.9 | 78 | 70.4 | 58.5 | 44.4 | 66.8 | — | — | |
| Baseline(ANN)Type=ANN, Backbone=ViT-S [17]2026.06 | 82.7 | 77.9 | 69.8 | 60.2 | 46 | 67.3 | — | — | |
| SpikeTAD(ANN)Type=ANN, Backbone=ViT-S [17]2026.06 | 82.6 | 77.7 | 69.7 | 60 | 45.6 | 67.1 | — | — | |
| ActionFormerBackbone=I3D2024.03 | 82.1 | 77.8 | 71 | 59.4 | 43.9 | 66.8 | — | — | |
| ActionFormerParams(M)=29.2, Backbone=I3D2024.09 | 82.1 | 77.8 | 71 | 59.4 | 43.9 | 66.8 | — | — | |
| ActionFormerType=ANN, Backbone=I3D [15]2026.06 | 82.1 | 77.8 | 71 | 59.4 | 43.9 | 66.8 | — | — | |
| SpikeTADType=ANN2SNN, Backbone=ViT-S [17], T (time-steps in the backbone)=162026.06 | 82.1 | 77.9 | 70.5 | 59.9 | 45.6 | 67.2 | — | — | |
| SpikeTADType=ANN2SNN, Backbone=ViT-S [17], T (time-steps in the backbone)=82026.06 | 81.4 | 76.6 | 70 | 60 | 44.8 | 66.5 | — | — | |
| Baseline_TADParams(M)=6.9, Backbone=ViT-B2024.09 | 80.6 | 74.2 | 66.3 | 53.1 | 37.3 | 62.3 | — | — | |
| Temporal2Seq_dataParams(M)=6.9, Backbone=ViT-B2024.09 | 80.3 | 74.2 | 67.7 | 55.2 | 40.3 | 63.5 | — | — | |
| Temporal2Seq_batchParams(M)=6.9, Backbone=ViT-B2024.09 | 79.9 | 74.6 | 67.4 | 56.4 | 41.9 | 64 | — | — | |
| ViT-TADType=ANN, Backbone=ViT-S [17]2026.06 | 79.8 | 75.2 | 68.4 | 56.4 | 41.7 | 64.3 | — | — | |
| ActionFormerParams(M)=29.2, Backbone=ViT-B2024.09 | 79.7 | 74.3 | 68.2 | 57.3 | 42.3 | 64.3 | — | — | |
| SpikeTADType=ANN2SNN, Backbone=ViT-S [17], T (time-steps in the backbone)=42026.06 | 79 | 74.6 | 67 | 57.2 | 42.3 | 64 | — | — | |
| ActionFormerType=ANN, Backbone=SlowFast-R50 [43]2026.06 | 78.7 | 73.3 | 65.2 | 54.6 | 39.7 | 62.3 | — | — | |
| Re2TALType=ANN, Backbone=Re2SlowFast-101 [13]2026.06 | 77.4 | 72.6 | 64.9 | 53.7 | 39 | 61.5 | — | — | |
| Re2TALType=ANN, Backbone=Re2VideoSwin-T [13]2026.06 | 77 | 71.5 | 62.4 | 49.7 | 36.3 | 59.4 | — | — | |
| TALLFormerParams(M)=15.0, Backbone=Swin-B2024.09 | 76 | — | 63.2 | — | 34.5 | 59.2 | — | — | |
| TALLFormerType=ANN, Backbone=VideoSwin-B [45]2026.06 | 76 | — | 63.2 | — | 34.5 | 59.2 | — | — | |
| BasicTADType=ANN, Backbone=SlowOnly-R50 [43]2026.06 | 75.5 | 70.8 | 63.5 | 50.9 | 37.4 | 59.6 | — | — | |
| TadTRFeature=I3D, single-network=true, end-to-end=false, focal loss=true, IBIF=true2021.06 | 74.8 | 69.1 | 60.1 | 46.6 | 32.8 | 56.7 | 195 | 1.07 | |
| Self-DETRBackbone=I3D2024.03 | 74.6 | 69.5 | 60 | 47.6 | 31.8 | 56.7 | — | — | |
| TadMLType=One-stage, Flow stream=true2022.06 | 73.29 | 69.73 | 62.53 | 53.36 | 39.6 | 59.7 | — | — | |
| Zhu et al.Backbone=I3D2024.03 | 72.1 | 65.9 | 57 | 44.2 | 28.5 | 53.5 | — | — | |
| TadTR-liteFeature=I3D, single-network=false, end-to-end=false, focal loss=true, IBIF=true2021.06 | 71.3 | 65.9 | 57 | 44.6 | 30.4 | 53.8 | 155 | 0.85 | |
| TadTRFeature=I3D, single-network=false, end-to-end=false, IBIF=true2021.06 | 70.3 | 64.3 | 55.7 | 44 | 30 | 52.9 | 195 | 1.07 | |
| RCLBackbone=I3D2024.03 | 70.1 | 62.3 | 52.9 | 42.7 | 30.7 | 51 | — | — | |
| DINOBackbone=I3D2024.03 | 69.8 | 63.1 | 53.7 | 41.5 | 26.4 | 50.9 | — | — | |
| ReActBackbone=TSN2024.03 | 69.2 | 65 | 57.1 | 47.8 | 35.6 | 55 | — | — | |
| MUSESFeature=I3D, single-network=false, end-to-end=false2021.06 | 68.9 | 64 | 56.9 | 46.3 | 31 | 53.4 | 2,101 | 34.1 | |
| MUSESBackbone=I3D2024.03 | 68.9 | 64 | 56.9 | 46.3 | 31 | 53.4 | — | — | |
| TadMLType=One-stage, Flow stream=false2022.06 | 68.78 | 64.66 | 56.61 | 45.4 | 31.88 | 53.46 | — | — | |
| TAGSBackbone=I3D2024.03 | 68.6 | 63.8 | 57 | 46.3 | 31.8 | 52.8 | — | — | |
| RTD-NetFeature=I3D, single-network=false, end-to-end=false2021.06 | 68.3 | 62.3 | 51.9 | 38.8 | 23.7 | 49 | — | — | |
| ContextLocType=Two-stage, Flow stream=false2022.06 | 68.3 | 63.8 | 54.3 | 41.8 | 26.2 | 50.88 | — | — | |
| ContextLocBackbone=I3D2024.03 | 68.3 | 63.8 | 54.3 | 41.8 | 26.2 | 50.9 | — | — | |
| RTD-NetBackbone=I3D2024.03 | 68.3 | 62.3 | 51.9 | 38.8 | 23.7 | 49 | — | — | |
| DCANBackbone=TSN2024.03 | 68.2 | 62.7 | 54.1 | 43.9 | 32.6 | 52.3 | — | — | |
| AFSDFeature=I3D, single-network=true, end-to-end=true2021.06 | 67.3 | 62.4 | 55.5 | 43.7 | 31.1 | 52 | 3,245 | 84.1 | |
| AFSDBackbone=I3D2024.03 | 67.3 | 62.4 | 55.5 | 43.7 | 31.1 | 52 | — | — | |
| AFSDParams(M)=14.2, Backbone=I3D2024.09 | 67.3 | 62.4 | 55.5 | 43.7 | 31.1 | 52 | — | — | |
| TadTRFeature=I3D, single-network=false, end-to-end=false2021.06 | 67.1 | 61.1 | 52 | 39.9 | 26.2 | 49.3 | 195 | 1.07 | |
| VSGNBackbone=TSN2024.03 | 66.7 | 60.4 | 52.4 | 41 | 30.4 | 50.2 | — | — | |
| VSGNParams(M)=8.4, Backbone=TSN2024.09 | 66.7 | 60.4 | 52.4 | 41 | 30.4 | 50.2 | — | — | |
| G-TAD + P-GCNFeature=I3D, single-network=false, end-to-end=false2021.06 | 66.4 | 60.4 | 51.6 | 37.6 | 22.9 | 47.8 | — | — | |
| G-TADParams(M)=6.1, Backbone=TSN2024.09 | 66.4 | 60.4 | 51.6 | 37.6 | 22.9 | 47.8 | — | — | |
| AGTFeature=I3D, single-network=true, end-to-end=true2021.06 | 65 | 58.1 | 50.2 | — | — | — | — | — | |
| P-GCN+Feature=I3D, single-network=false, end-to-end=false2021.06 | 64.9 | 59 | 49.4 | 36.7 | 22.6 | 46.5 | 7,298 | 4.4 | |
| BMN-CSABackbone=TSN2024.03 | 64.4 | 58 | 49.2 | 38.2 | 27.8 | 47.7 | — | — | |
| PCG-TALFeature=I3D, single-network=false, end-to-end=false2021.06 | 64.2 | 57.3 | 48.3 | — | — | — | — | — | |
| P-GCNFeature=I3D, single-network=false, end-to-end=false2021.06 | 63.6 | 57.8 | 49.1 | — | — | — | 7,298 | 4.4 | |
| GCNType=Two-stage, Flow stream=false2022.06 | 63.6 | 57.8 | 49.1 | — | — | — | — | — | |
| TadTRType=One-stage, Flow stream=false2022.06 | 62.4 | 57.4 | 49.2 | 37.8 | 26.3 | 46.6 | — | — | |
| TadTRBackbone=I3D2024.03 | 62.4 | 57.4 | 49.2 | 37.8 | 26.3 | 46.6 | — | — | |
| TadTRParams(M)=8.6, Backbone=I3D2024.09 | 62.4 | 57.4 | 49.2 | 37.8 | 26.3 | 46.6 | — | — | |
| TadTRType=ANN, Backbone=I3D [15]2026.06 | 62.4 | 57.4 | 49.2 | 37.8 | 26.3 | 46.6 | — | — | |
| TCA-NetBackbone=TSN2024.03 | 60.6 | 53.2 | 44.6 | 36.8 | 26.7 | 44.3 | — | — | |
| BSN++Type=Two-stage, Flow stream=false2022.06 | 59.9 | 45.9 | 41.3 | 31.9 | 22.8 | 40.36 | — | — | |
| G-TADFeature=I3D, single-network=false, end-to-end=false, result_source=copied from [36]2021.06 | 58.7 | 52.7 | 44.9 | 33.6 | 23.8 | 42.7 | — | — | |
| G-TADType=Two-stage, Flow stream=false2022.06 | 58.7 | 52.7 | 44.9 | 33.6 | 23.8 | 42.7 | — | — | |
| A2NetFeature=I3D, single-network=true, end-to-end=true2021.06 | 58.6 | 54.1 | 45.5 | 32.5 | 17.2 | 41.6 | 1,554 | 30.4 | |
| A2NetType=One-stage, Flow stream=false2022.06 | 58.6 | 54.1 | 45.5 | 32.5 | 17.2 | 41.6 | — | — | |
| PBRNetType=Two-stage2022.06 | 58.5 | 54.6 | 51.3 | 41.8 | 29.5 | — | — | — | |
| RTD-NetParams(M)=14.0, Backbone=I3D2024.09 | 58.5 | 53.1 | 45.1 | 36.4 | 25 | 43.6 | — | — | |
| DBGType=Two-stage, Flow stream=false2022.06 | 57.8 | 49.4 | 39.8 | 30.2 | 21.7 | 39.78 | — | — | |
| G-TADType=One-stage2022.06 | 57.8 | 47.2 | 38.8 | — | — | — | — | — | |
| BC-GNNFeature=TS, single-network=false, end-to-end=false2021.06 | 57.1 | 49.1 | 40.4 | 31.2 | 23.1 | 40.2 | — | — | |
| BC-GNNBackbone=TSN2024.03 | 57.1 | 49.1 | 40.4 | 31.2 | 23.1 | 40.2 | — | — | |
| BMNFeature=I3D, single-network=false, end-to-end=false, result_source=copied from [36]2021.06 | 56.4 | 47.9 | 39.2 | 30.2 | 21.2 | 39 | — | — | |
| BMNFeature=TS, single-network=false, end-to-end=false2021.06 | 56 | 47.4 | 38.8 | 29.7 | 20.5 | 38.5 | — | — | |
| BMNType=Two-stage, Flow stream=false2022.06 | 56 | 47.4 | 38.8 | 29.7 | 20.5 | 38.48 | — | — | |
| BMNBackbone=TSN2024.03 | 56 | 47.4 | 38.8 | 29.7 | 20.5 | 38.5 | — | — | |
| BMNType=ANN, Backbone=TSN [16]2026.06 | 56 | 47.4 | 38.8 | 29.7 | 20.5 | 38.5 | — | — | |
| G-TADFeature=TS, single-network=false, end-to-end=false2021.06 | 54.5 | 47.6 | 40.2 | 30.8 | 23.4 | 39.3 | — | — | |
| G-TADBackbone=TSN2024.03 | 54.5 | 47.6 | 40.2 | 30.8 | 23.4 | 39.3 | — | — | |
| MGGFeature=TS, single-network=false, end-to-end=false2021.06 | 53.9 | 46.8 | 37.4 | 29.5 | 21.3 | 37.8 | — | — | |
| MRFeature=I3D, single-network=false, end-to-end=false2021.06 | 53.9 | 50.7 | 45.4 | 38 | 28.5 | 43.3 | — | — | |
| MGGType=Two-stage, Flow stream=false2022.06 | 53.9 | 46.8 | 37.4 | 29.5 | 21.3 | 37.78 | — | — | |
| MRType=Two-stage, Flow stream=false2022.06 | 53.9 | 50.7 | 45.4 | 38 | 28.5 | 43.3 | — | — | |
| TAL-MRBackbone=I3D2024.03 | 53.9 | 50.7 | 45.4 | 38 | 28.5 | 43.3 | — | — | |
| BSNFeature=TS, single-network=false, end-to-end=false2021.06 | 53.5 | 45 | 36.9 | 28.4 | 20 | 36.8 | — | — | |
| TAL-NetFeature=I3D, single-network=true, end-to-end=false2021.06 | 53.2 | 48.5 | 42.8 | 33.8 | 20.8 | 39.8 | — | — | |
| TAL-NetType=Two-stage, Flow stream=false2022.06 | 53.2 | 48.5 | 42.8 | 33.8 | 20.8 | 39.8 | — | — | |
| SSNFeature=TS, single-network=false, end-to-end=false2021.06 | 51.9 | 41 | 29.8 | — | — | — | — | — | |
| PALPre-training dataset=K400, Training epochs=100, TAD detector framework=G-TAD2022.03 | 46.8 | 40.3 | 30.8 | 19.3 | 10.9 | — | — | — | |
| R-C3DFeature=C3D, single-network=true, end-to-end=true2021.06 | 44.8 | 35.6 | 28.9 | — | — | — | — | — | |
| R-C3DType=Two-stage, Flow stream=false2022.06 | 44.8 | 35.6 | 28.9 | — | — | — | — | — | |
| TACPre-training dataset=K400, Training epochs=100, TAD detector framework=G-TAD2022.03 | 44.6 | 37.3 | 29.5 | 18.8 | 9.5 | — | — | — | |
| TURN-TAPType=Two-stage, Flow stream=false2022.06 | 44.1 | 34.9 | 25.6 | — | — | — | — | — | |
| SSADFeature=TS, single-network=true, end-to-end=false2021.06 | 43 | 35 | 24.6 | — | — | — | — | — | |
| MoCo-v2Pre-training dataset=K400, Training epochs=100, TAD detector framework=G-TAD2022.03 | 41.5 | 34.1 | 25.8 | 17.3 | 7.9 | — | — | — | |
| CDCType=Two-stage, Flow stream=false2022.06 | 40.1 | 29.4 | 23.3 | 13.1 | 7.9 | 20.76 | — | — | |
| Yuan et al.Feature=TS, single-network=false, end-to-end=false2021.06 | 36.5 | 27.8 | 17.8 | — | — | — | — | — | |
| Yeung et al.Feature=VGG16, single-network=true, end-to-end=false2021.06 | 36 | 26.4 | 17.1 | — | — | — | — | — |