Transfer from: http: //www.cvchina.info/2011/04/15/facel/
These days individuals will use OpenCV face detection, the code can be found anywhere online that really matter two lines of code. But if you want to find complex and fly in the face of computer vision related to the code it is really very little of. Complex: at least not AdaBoost + CAMSHFIT on the finished thing, fly: Algorithms do not much fancy, work on the line (in fact, many times it is difficult to do), and the code is written in neat, have documentation comments. Currently I know there are one or two such open source projects, very strange exposure is very low, then come to share.
This is FaceL, the project home page here , author David S. Bolme and J. Ross Beveridge from Colorado State University. FaceL is based on a video capture OpenCV python interface implementations, face tagging (Labeling), training (Training), and recognition (Classification) of one-stop software. From the official video , the effect is very tricky! But only a few hundred views. . . I really wonder, there are more than 8,000 page views in attendance do I get weak weak face recognition demo. . . Oops
label can be a person's identity (Identity), it can also be expression (expression), or you can go without a label with sunglasses and sunglasses. How to say it is just a label, classifiers do not care what the specific meaning. FaceL LibSVM of using a Python interface directly integrated and real-time online training.
To achieve the above functions, there are ready-made algorithms. Looks FaceL credit is to save up and everything, but this has been terrific, you can save a lot of time Computer Vision practitioners developers. But more Niubi is FaceL also implements eye tracker called ASEF, very tricky accurate, and soon, is again convolution. Algorithm from the author's own CVPR on paper in 2009: Average of Synthetic Exact fi lters . I think this stuff is useful, it is to port C code to run on the iPhone is unambiguous.
Project was developed on a Mac, but windows and Linux can run. To sum up FaceL is enough to power!
Reproduced in: https: //my.oschina.net/dake/blog/196838