|
Is SVM suitable for face recognition? If I have number of person = number of classes, so I have a lot of classes and small number of samples per class. EDIT: found interesting paper comparing PCA and SVM. |
|
Hi, ofcourse SVM is suitable for face recognition :), you can use one vs. all multi-classification paradigm. However, since you have one sample per class, you must have discriminating features, but I recommend you try get many more samples still. To get these features you would need to use feature extractors such as DCT, FFT, or RBM. PCA is common but there are better feature extractors. Here is a reference that might help, |
