Action Recognition on THUMOS 14 (test)
84.7mAPR(2+1)D & DSN (M = 1)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| R(2+1)D & DSN (M = 1)Input=RGB+FLOW, Clips=62020.06 | 84.7 | — | |
| R(2+1)D & DSN (M = 1)Input=RGB, Clips=62020.06 | 83.1 | — | |
| R(2+1)DInput=RGB+FLOW, Clips=202020.06 | 82.6 | — | |
| UntrimmedNet (soft)Input=RGB+FLOW, Clips=ALL2020.06 | 82.2 | — | |
| R(2+1)DInput=RGB, Clips=202020.06 | 81.9 | — | |
| UntrimmedNet (hard)Input=RGB+FLOW, Clips=ALL2020.06 | 81.2 | — | |
| TSNInput=RGB+FLOW, Clips=ALL2020.06 | 80.1 | — | |
| R(2+1)D & DSN (M = 1)Input=FLOW, Clips=62020.06 | 78.4 | — | |
| R3D-RGB & DSN (M = 6)Input=RGB, Clips=62020.06 | 75.2 | — | |
| R3D-RGB & TSN (6 seg)Input=RGB, Clips=202020.06 | 73.6 | — | |
| R(2+1)DInput=FLOW, Clips=202020.06 | 73.5 | — | |
| Objects+MotionInput=RGB+HOG+HOF+MBH, Clips=ALL2020.06 | 71.6 | — | |
| Jain et al.Input=HOG+HOF+MBH, Clips=ALL2020.06 | 71 | — | |
| Two StreamInput=RGB+FLOW, Clips=252020.06 | 66.1 | — | |
| MotionInput=HOG+HOF+MBH, Clips=ALL2020.06 | 63.1 | — | |
| ELMInput=HOG+HOF+MBH, Clips=ALL2020.06 | 62.3 | — | |
| iDT+CNNInput=RGB+iDT, Clips=ALL2020.06 | 62 | — | |
| RGB+EMVInput=RGB+FLOW2020.06 | 61.5 | — | |
| CDC2017.03 | 44.4 | — | |
| Conv & De-convUpsampling method=Separate convolutional and de-convolutional layers2017.03 | 41.7 | — | |
| MultiLSTMMechanism=Temporal attention2017.03 | 41.3 | — | |
| LSTMHidden Units=512, Feature Layer=VGG CNN FC72017.03 | 39.3 | — | |
| CDCTemporal Modeling=Fixed 3D ConvNets2017.03 | 37.4 | — | |
| C3D + LinearInterpUpsampling method=Linear interpolation, Base Model=C3D2017.03 | 37 | — | |
| Two-stream CNNInputs=Pixel and Optical Flow2017.03 | 36.2 | — | |
| Single-frame CNNBackbone=16-layer VGG CNN2017.03 | 34.7 | — | |
| EMV+RGBSupervision Level=Strong supervision2017.03 | — | 61.5 | |
| EMV+RGB2017.05 | — | 61.5 | |
| iDT+FVSupervision Level=Strong supervision2017.03 | — | 63.1 | |
| iDT+FV2017.05 | — | 63.1 | |
| object+motion2017.05 | — | 71.6 | |
| Objects+MotionSupervision Level=Strong supervision2017.03 | — | 71.6 | |
| TSNNumber of segments=3, Supervision Level=Strong supervision2017.03 | — | 78.5 | |
| TSNNumber of segments=32017.05 | — | 78.8 | |
| TSNNumber of segments=72017.05 | — | 80.1 | |
| Two StreamSupervision Level=Strong supervision2017.03 | — | 66.1 | |
| Two Stream2017.05 | — | 66.1 | |
| UntrimmedNetSelection Module=hard, Supervision Level=Weak supervision2017.03 | — | 81.2 | |
| UntrimmedNetSelection Module=soft, Supervision Level=Weak supervision2017.03 | — | 82.2 |