Hello, I'm looking for an example of face recognition on videos (like the demo for an image). The only examples I see are with a camera. Do you have a solution? Thank you.
I tested it and it sample is not useful. We want to face detection and recognation face for example in live video if show my face then in top of my face show 'Ali Abbasi' and if in live video show your face them a top of face show 'Nigel'. We want to face recognation and face detection is not complete. I am sure it is very useful for many developers. Please see in the attachment file.
For my part I would like to use a film to capture several faces. The examples you provide seem to work only with a computer camera and I don't see how to modify the sources to adapt them to a video. Thank you for your help. Best Regards
Thank you, I downloaded new version of it demo and test it. -I add detected face name 'Ali' and other face to 'Bob'. When 'Ali' and 'Bob' in one time show in viedo then all faces name show 'Bob'. Also again 'Ali' show to video then name 'Bob'. Allwas show last name saved!
-Also if face name show above of face box then it is better.(similar my above pictures of 'Mike' and ' Michael '.
-I am sure which you can make a good face recognation for us!
Every time you click the "Add Detected Face..." (even if it already shows a name) the image is added to the face database and improves the accuracy. You can confirm this by clicking the "View Face Database" button.
I tested it and accuracy for recognation not good. Please you will test with 3 or over persons tested then you can find confuse for face recognation beaten persons faces.
I tested again and below is results: -face detection is ok which I think it is result of OpenCV. -Problem is face recognation and if face recognated for example 'Ali' then show in above face name 'Ali' and if do not recognated then do not show face name (show face name 'unknown. Main problem is if face not recognated then show pervios face name and if do not recognated face then show 'unknown'. If you fix it then I think it is ok.
Are you able to show me a demo where openCV facial recognition works better than what we are getting?
I think the OpenCV methods are quite sensitive to lighting and other effects, and don't compare to nearal network methods.
The demo uses the Eigenfaces algorithm. You could try testing the other alogorithms too: - Fisherfaces - Local Binary Patterns Histograms (LBPH)
confidence is one of the parameters of the detect method. You can use that to determine how confident OpenCV is in its detection. If detection fails the returned label will be blank, but in my testing it always returns a label.