Hello,
I'm trying to use the FaceRecognition class to recognize faces in video, but it seems that it has accuracy of about 20%-30%. These algorithms supposed to have 97% of accuracy, so i might be doing something wrong.
I used the example here: http://docs.opencv.org/trunk/modules/contrib/doc/facerec/tutorial/facerec_video_recognition.html
I tried the Fisherfaces and also the Eigenfaces algorithms. For training I used 3 images of each person (only 3 persons). The pictures were taken with an iPhone camera (good resolution) with good lighting conditions and with a white-yellow wall behind the person. I also aligned the pictures with the alignment python script. The dimension of the pictures are 300X300.
The prediction is made on my laptop (standard camera) on several lighting conditions, and most of the times the prediction is wrong.
What did I miss?
Thank you, Edi.