Research Article
Fusing Two Kinds of Virtual Samples for Small Sample Face Recognition
Table 3
Classification error rates of different algorithms on the Yale database.
| The number of training samples | 1 | 2 | 3 |
| CRC-original | 0.3733 | 0.1556 | 0.1083 | CRC-mirror face | 0.2800 | 0.1259 | 0.0333 | CRC-symmetrical face | 0.3067 | 0.1037 | 0.0500 | CRC-the proposed scheme | 0.2200 | 0.0889 | 0.0250 |
| INNC | 0.3733 | 0.2296 | 0.1833 | INNC-mirror face | 0.2800 | 0.1852 | 0.1250 | INNC-symmetrical face | 0.3067 | 0.2000 | 0.1583 | INNC-the proposed scheme | 0.2200 | 0.1630 | 0.1000 |
| SFRFR | 0.3733 | 0.2148 | 0.1417 | SFRFR-mirror face only | 0.2800 | 0.2000 | 0.0833 | SFRFR-symmetrical face only | 0.3067 | 0.1778 | 0.1167 | SFRFR-the proposed scheme | 0.2200 | 0.1333 | 0.0417 |
| TPTSR | 0.3733 | 0.1926 | 0.1417 | TPTSR-mirror face only | 0.2800 | 0.1259 | 0.0333 | TPTSR-symmetrical face only | 0.3067 | 0.1037 | 0.0500 | TPTSR-the proposed scheme | 0.2200 | 0.0889 | 0.0250 |
|
|