Research Article
Fusing Two Kinds of Virtual Samples for Small Sample Face Recognition
Table 5
Classification error rates of different algorithms on the FLW database.
| The number of training samples | 1 | 2 | 3 |
| CRC-original | 0.8437 | 0.7660 | 0.7641 | CRC-mirror face | 0.8282 | 0.7485 | 0.7276 | CRC-symmetrical face | 0.8359 | 0.7558 | 0.7359 | CRC-the proposed scheme | 0.8178 | 0.7485 | 0.7209 |
| INNC | 0.8437 | 0.7718 | 0.7674 | INNC-mirror face | 0.8282 | 0.7747 | 0.7425 | INNC-symmetrical face | 0.8359 | 0.7631 | 0.7425 | INNC-the proposed scheme | 0.8178 | 0.7631 | 0.7342 |
| SFRFR | 0.8437 | 0.7922 | 0.7724 | SFRFR-mirror face only | 0.8282 | 0.7631 | 0.7392 | SFRFR-symmetrical face only | 0.8359 | 0.7645 | 0.7409 | SFRFR-the proposed scheme | 0.8178 | 0.7456 | 0.7209 |
| TPTSR | 0.8346 | 0.7442 | 0.7276 | TPTSR-mirror face only | 0.8217 | 0.7413 | 0.7010 | TPTSR-symmetrical face only | 0.8140 | 0.7311 | 0.7010 | TPTSR-the proposed scheme | 0.8127 | 0.7311 | 0.6910 |
|
|