hi i am happy i found this project looking for detection of glasses on a face. i was hoping i can just grab your glass_sp.dat model and start detecting with dlib. i got some results with this but the detection seems to work at least a little bit but is very inaccurate. i do not use your code though.
i tried to crop the pic manually to just the eyes area (like you seem to do in your code) and then detect. the result is not accurate and also has big errors.
i wonder what i can do different to get good results.
just to be clear i am able to use shape_predictor_68_face_landmarks.dat in my project and detect face landmarks very accurate. i just hoped the glas detection would be just as easy, with just use your model instead.
i also wonder what training data you used? what did your detection results look like? do you have any results to show?
this is my result from a cropped image (i think in your code you use the svm to find the glasses first and then do landmark detection on that area only)
alos tried with glasses only but still not so good.
hi i am happy i found this project looking for detection of glasses on a face. i was hoping i can just grab your glass_sp.dat model and start detecting with dlib. i got some results with this but the detection seems to work at least a little bit but is very inaccurate. i do not use your code though.
i tried to crop the pic manually to just the eyes area (like you seem to do in your code) and then detect. the result is not accurate and also has big errors.
i wonder what i can do different to get good results.
just to be clear i am able to use shape_predictor_68_face_landmarks.dat in my project and detect face landmarks very accurate. i just hoped the glas detection would be just as easy, with just use your model instead.
i also wonder what training data you used? what did your detection results look like? do you have any results to show?
thanks you