Table 3 ADD(-S) metric of multi object 6D pose estimation using a single model on the Occlusion dataset. Symmetric objects are marked with *.
From: MagicCubePose, A more comprehensive 6D pose estimation network
Method | PoseCNN | PVNet | RNNPose | EfficientPose27(\(\varphi\)=0) | 27(\(\varphi\)=3) | Our(\(\varphi\)=0) | Our(\(\varphi\)=1) |
|---|---|---|---|---|---|---|---|
Ape | 9.60 | 15.8 | 37.18 | 56.57 | 59.39 | 56.73 | 59.34 |
Can | 45.2 | 63,3 | 88.07 | 91.12 | 93.27 | 92.21 | 94.44 |
Cat | 0.93 | 16.7 | 29.15 | 68.58 | 79.78 | 68.59 | 80.13 |
Driller | 41.4 | 65.7 | 88.14 | 95.64 | 97.77 | 95.67 | 97.77 |
Duck | 19.6 | 25.2 | 49.17 | 65.31 | 72.71 | 66.54 | 73.32 |
Eggbox * | 22.0 | 50.2 | 66.98 | 93.46 | 96.18 | 95.33 | 96.34 |
Glue * | 38.5 | 49.6 | 63.79 | 85.15 | 90.80 | 85.45 | 90.81 |
Holepuncher | 22.1 | 39.7 | 62.76 | 76.53 | 81.95 | 76.61 | 81.97 |
Average | 24.9 | 40.8 | 60.65 | 79.04 | 83.98 | 79.64 | 84.27 |
