Extract Feature Detection from OpenCV Facial Recognition Algorithm
Hello,
I am aware that OpenCV has algorithms like Local Binary Patterns and Eigenfaces for facial detection as described here: http://docs.opencv.org/2.4/modules/co...
I understand that these algorithms compute features for the faces, and would like to exact these features.
These features are like the ones used in machine learning algorithms, not parts of the face.
Is it possible to extract features from the OpenCV facial recognition algorithms?
For example, in the link it talks about a 22-dimensional feature vector that was used once.
Thanks
Chapter 6 of OpenCV 3 Blueprints talks about biometric identification and also discusses various parts of the face recognition interface. Might be interesting to take a look. About the feature vector, it is generated using reprojection. For example, the eigen representations can be retrieved using this part of the code.
As far as the part of the code, I think you have to first train the model in order to extract the eigen representations. I don't want to train the model. I just want to get the feature vectors.
Well as suggested in this sample of the book I mentioned, for LBPH face recognizers you can use the ELBP functionality to get the feature vectors, without training a recognizer model itself. I guess that for eigen and fisher representations, you will have to crack open the code and look for the part where the features are generated, respectively here and here. Then you will need to copy the respective code into a seperate function for your use.
i know, it's matlab, but still a nice overview, what has been tried with "textrons"