How to use Catalano framework for facial feature extraction? By the way, we use Java

Question

The goal of the program was to create a real-time emotional recognition system. We will be using opencv for face detection. Is it possible to use Catalano Framework for extracting the facial features from the picture we get from face detection? If possible then how? Also, How do you use fuzzy logic as an emotional classifier after getting the extracted facial features? Thanks in advance.

Answer 1

Yes, It's possible to extract the facial features, there is several texture algorithms (i.e LBP, LBPu2, ILBP, CS-LBP, etc...), contains 14 LBP variants to work, and some pre-processing photometric illuminant filters like Retinex, Difference of Gaussians (DoG), Tan-Triggs Normalization, Gradient Faces, Retina Model. Also there is histogram approach like Spatial Histogram and Spatial Pyramid Histogram.

This image shows my approach using the framework. The approach is similar like the D. MATURANA.

First face, is the original image from the ORL database. Second face, I had used the Self Quocient Image (I'm currently implementing this). Third face, is the Spatial Histogram (6x6) representation.

So I compute ILBP for each block. I can extract 66511 = 18396 features.

With SVM using Chi Square kernel and leave-one-out, I can achieve 99.25% accuracy.

*Disclaimer: I'm the author of the Catalano Framework