Table 6 Sensitivity analysis for spatial stream parameters on PoseTrack dataset.
From: Learning spatio-temporal context for basketball action pose estimation with a multi-stream network
Conv layers | Filter size | Num filters | Head AP (%) | Shoulder AP (%) | MAP (%) |
|---|---|---|---|---|---|
3 Layers | 3x3 | 64 | 81.2 | 84.3 | 79.1 |
3 Layers | 5x5 | 64 | 81.7 | 84.8 | 79.6 |
5 Layers | 3x3 | 64 | 82.1 | 85.3 | 80.2 |
5 Layers | 3x3 | 128 | 82.5 | 85.6 | 80.7 |
5 Layers | 5x5 | 128 | 82.9 | 85.9 | 81.0 |
7 Layers | 3x3 | 128 | 83.0 | 86.0 | 81.2 |
