Research Article
Pose-Invariant Face Recognition via RGB-D Images
Table 1
Recognition rates for different pose rotations on the Bosphorus database.
| Pose |
PGM [16] |
PGDP [17] | Proposed method | Depth | Texture | Fusion |
| +10° right | 92.3% | 96.1% | 90.0% | 100% | 99.9% | +20° right | 88.6% | 91.4% | 89.5% | 99.7% | 99.9% | +30° right | 80.0% | 80.1% | 89.3% | 99.6% | 99.8% | +45° right | 39.0% | 36.2% | 88.8% | 97.7% | 98.7% | −45° left | 38.1% | 37.1% | 88.6% | 98.1% | 98.6% | +90° right | N/A | N/A | 84.2% | 91.4% | 95.1% | −90° left | N/A | N/A | 84.5% | 92.3% | 95.7% | Upwards (slight) | 87.6% | 96.2% | 88.9% | 93.4% | 97.3% | Upwards (strong) | 79.0% | 91.4% | 87.1% | 92.0% | 96.6% | Downwards (slight) | 87.6% | 86.7% | 87.1% | 93.0% | 97.8% | Downwards (strong) | 69.5% | 70.0% | 85.4% | 92.7% | 94.8% | Right-upwards | 63.8% | 64.7% | 85.6% | 91.2% | 95.6% | Right-downwards | 34.3% | 35.2% | 84.3% | 89.0% | 92.5% | Average | 69.1% | 71.4% | 87.1% | 93.4% | 96.7% |
|
|
N/A denotes that the measure is not available in the paper where the method was proposed (the same for the other tables in this paper).
|