Download Face Recognition dataset from: http://vision.ucsd.edu/content/yale-face-database
center-light, w/glasses, happy, left-light, sleepy, surprised, wink
w/no glasses, normal, right-light, sad
Accuracy | One-vs-All | Multi-Class |
---|---|---|
PCA | 75.0% | 75.0% |
LDA | 86.7% | 90.0% |
PCA+SVM* | 88.3% | N/A |
- 95% variance are kept after dimensionality reduction, which degrades SVM performance