TABLE 6.
Performance of non-rigid SFM methods on synthetic and motion capture data. For the related PTA and CSF-Bnr methods, factorization rank is also indicated by the value of K in parenthesis.
| Dataset | XCK | EM-PPCA | MP | PTA | CSF-Bnr | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| eR | e 3D | eR | e 3D | eR | e 3D | eR | e3D (K) | eR | e3D (K) | initial e3D (K) | |
| Drink | .3359 | 3.5186 | .2906 | .3393 | .2859 | .4604 | .0058 | .0250 (13) | .0055 | .0223 (6) | .0854 (6) |
| Pick-up | .4687 | 3.3721 | .4277 | .5822 | .2506 | .4332 | .1549 | .2369 (12) | .1546 | .2301 (6) | .2685 (6) |
| Yoga | 1.2014 | 7.4935 | .8089 | .8097 | .8711 | .8039 | .1059 | .1625 (11) | .1021 | .1467 (7) | .1528 (7) |
| Stretch | .9489 | 4.2415 | .7594 | 1.1111 | .8174 | .8549 | .0549 | .1088 (12) | .0489 | .0710 (8) | .0966 (8) |
| Dance | – | 2.9962 | – | .9839 | – | .2639 | – | .2958 (5) | – | .2705 (2) | .3259 (2) |
| Face1 | – | – | – | .0434 | – | .0734 | – | .1247 (3) | – | .0637 (5) | .1487 (3) |
| Face2 | – | – | – | .0329 | – | .0357 | – | .0444 (5) | – | .0363 (3) | .0451 (3) |
| Shark | – | – | – | .0501 | – | .1571 | – | .1796 (9) | – | .0081 (3) | .3195 (3) |
| Walking | – | – | – | .4917 | – | .5607 | – | .3954 (2) | – | .1863 (2) | .6823 (2) |