Online Action Detection on THUMOS 14
71.6Mean F-APMAT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MATArch=Trans, Pre-trained Features=Kinetics2023.08 | 71.6 | — | — | — | |
| TeSTraArch=Trans, Pre-trained Features=Kinetics2023.08 | 71.2 | — | — | — | |
| GateHUBArch=Trans, Pre-trained Features=Kinetics2023.08 | 70.7 | — | — | — | |
| MATArch=Trans, Pre-trained Features=ANet2023.08 | 70.4 | — | — | — | |
| LSTRArch=Trans, Pre-trained Features=Kinetics2023.08 | 69.5 | — | — | — | |
| GateHUBBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 69.1 | — | — | — | |
| GateHUBArch=Trans, Pre-trained Features=ANet2023.08 | 69.1 | — | — | — | |
| TeSTraArch=Trans, Pre-trained Features=ANet2023.08 | 68.2 | — | — | — | |
| WOADArch=RNN, Pre-trained Features=Kinetics2023.08 | 67.1 | — | — | — | |
| ColarArch=Trans, Pre-trained Features=Kinetics2023.08 | 66.9 | — | — | — | |
| LSTRBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 65.3 | — | — | — | |
| LSTRArch=Trans, Pre-trained Features=ANet2023.08 | 65.3 | — | — | — | |
| OadTRArch=Trans, Pre-trained Features=Kinetics2023.08 | 65.2 | — | — | — | |
| PKDArch=CNN, Pre-trained Features=Kinetics2023.08 | 64.5 | — | — | — | |
| TRNArch=RNN, Pre-trained Features=Kinetics2023.08 | 62.1 | — | — | — | |
| IDNArch=RNN, Pre-trained Features=Kinetics2023.08 | 60.3 | — | — | — | |
| ColarArch=Trans, Pre-trained Features=ANet2023.08 | 59.4 | — | — | — | |
| OadTRBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 58.3 | — | — | — | |
| OadTRArch=Trans, Pre-trained Features=ANet2023.08 | 58.3 | — | — | — | |
| TFNBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 55.7 | — | — | — | |
| LAPBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 53.3 | — | — | — | |
| LAPArch=RNN, Pre-trained Features=ANet2023.08 | 53.3 | — | — | — | |
| FATSBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 51.6 | — | — | — | |
| IDNBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 50 | — | — | — | |
| IDNArch=RNN, Pre-trained Features=ANet2023.08 | 50 | — | — | — | |
| TRNBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 47.2 | — | — | — | |
| TRNArch=RNN, Pre-trained Features=ANet2023.08 | 47.2 | — | — | — | |
| REDBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 45.3 | — | — | — | |
| REDArch=RNN, Pre-trained Features=ANet2023.08 | 45.3 | — | — | — | |
| CDCBackbone=TSN, Pre-training Dataset=ActivityNet, Input Modality=RGB + Optical Flow2022.06 | 44.4 | — | — | — | |
| CDCArch=CNN, Pre-trained Features=ANet2023.08 | 44.4 | — | — | — | |
| WOADSupervision=S2020.06 | 0.671 | — | — | — | |
| WOADSupervision=V+50%S2020.06 | 0.626 | — | — | — | |
| WOADSupervision=V+30%S2020.06 | 0.593 | — | — | — | |
| WOADSupervision=V+10%S2020.06 | 0.55 | — | — | — | |
| WOADSupervision=V2020.06 | 0.544 | — | — | — | |
| TRNSupervision=S2020.06 | 0.51 | — | — | — | |
| CMeRT2026.03 | — | 78.2 | 48.9 | 34.6 | |
| DiscBackbone=Llama3.2-1B-Instruct2026.03 | — | 78.5 | 57.8 | 40.1 | |
| GADBackbone=Llama3.2-1B-Instruct2026.03 | — | 78.8 | 58.1 | 40.2 | |
| GenBackbone=Llama3.2-1B-Instruct2026.03 | — | 78.6 | 56.9 | 38.8 | |
| MAT2026.03 | — | 78.4 | 49.4 | 34.2 | |
| Testra2026.03 | — | 76.9 | 43 | 32 |