Open cesarsouza opened 7 years ago
Cool!
I might put my own collected assembled codes of face recognition from Accord.Net Framework with BOW, SVM and HMM...
Failed not perfected...needs to be done by a pro or an expert or intelligent person than I am...
lets wait and see who's first to contribute to start directly :D
@cesarsouza here's what I got, I hope it could help you even better its from you, lol I hope it more be easier for you to build this face recognizer function (from Accord.Net)
and @ANYONE, I hope you share your code experiment too.
Thank a lot @blaisexen, I am sure your code will serve as a good start!
Also just to leave as a reminder: Once this is done, this StackOverflow question should be answered as well.
@cesarsouza Glad I could help, and as long as I know that code setting recognizes cards perfectly from Ace to King using SVM even small images of cards to Large. :dancer:
and I hope you can build Accord.Net FaceRecognizers using msvm/hmm/backpropagation_neural_network/ or any that will surely recognized faces with label of "Unknown" and "Match"!
and surely a lot of software integrators/developers will use it and donate whatever the outcome of every project.
But as Far as I know that is very hard!
Good Luck making Accord.Net a Hardcore face Recognizer :dagger:
@cesarsouza
hi,
could it be possible that automation of face recognizers will be done in the next released?
Now I know it's so hard to make a face recognition system.
I am so waiting for this as well :(
I don't think this will come out!
Could you add a simple demo in it?
Thx for your awesome accord.net code!
The framework currently offers sample applications for face detection and tracking, but it does not offer samples for face recognition yet (i.e. being able to distinguish to which person each face corresponds).
Ideally, the sample application should be able to register a few faces from a video stream using the Viola-Jones face detector. Then, the user would have been able to separate each of the faces into separate classes (in a very similar way to the Dynamic Time Warping SVMs or Mouse Gesture Recognition sample application. After that, the user would have been able to train a SVM or equivalent classifier to recognize each of the faces in the created dataset. Finally, it should be possible to use the classifier to recognize new faces in new frames of incoming video.
This should help with #1018, #801, and possibly improve #724, #458.