Prebuild trainer for video face recognition processing in opencv c++ -

Currently I'm trying on gender validation / classification using OpenCV and Fisherfolus algorithms. Right now, I am training my program with 2500+ images with different programs.

The problem is that I take 35 minutes to execute my file. Every time I execute my file, I have to train my program again. Which will process up to 2500+ images and take the video to the Face recognition program for only 35 minutes.

My Execution Line in Terminal

  ./ main haarcascade_frontalface_alt.xml image_to_train.txt 0  

./ will be the main object < / P>

image_to_train.txt, where there are 2500 images

0 for webcams

Anyway the .txt file is pre-built on iamges, so I I can not train them every time I was thinking about a database, but then a database usually runs a single image processing through the whole database Land (brute-force concept).

Program training code These are the two lines in the entire video recognition process, images are passed in these functions with an execution time of once for 2500 images. Showing in OpenCV documentation I still do not think about how to remove data from them and how I can do them to speed up execution. A prebold instrument trainer.

  Ptr < FaceRecognizer & gt; Model = createFisherFaceRecognizer (); Model-> train (picture, label);