Skip to main content

Table 1 The joint discriminator networks classification accuracy on synthetic datasets of respective joint rotations

From: Learning 3D joint constraints from vision-based motion capture datasets

Joint name [16] Accuracy (%) False positive (%) False negative (%) Optimal hidden layer
Right up leg 98.7 0.3 1.0 8
Right foot 97.7 1 1.3 8
Left up leg 98.9 0.3 0.8 8
Left foot 98.2 0.6 1.2 8
Spine 99.4 0.1 0.5 16
Spine1 99.4 0 0.6 16
Neck 94.4 2.3 3.3 8
Head 97.7 0.9 1.4 8
Left shoulder 97.9 0.7 1.5 16
Left arm 96.8 0.5 2.7 8
Left hand 94.7 2.1 3.3 8
Right shoulder 95.1 1.5 3.3 16
Right arm 95.5 1.63 2.9 8
Right hand 95 2 3 8