Open arasharchor opened 7 years ago
I'm not going to do it. But maybe someone will contribute :)
@smajida , why not use shape_predictor for it? you simply need to have about 1000 images with human eyes, with pupil size about 3-10 pixels and train the model with the code from examples directory
@e-fominov good idea. Do you know a sample dataset for this? I found this https://www.mpi-inf.mpg.de/departments/computer-vision-and-multimodal-computing/research/gaze-based-human-computer-interaction/appearance-based-gaze-estimation-in-the-wild-mpiigaze/
and this one
http://www.cs.columbia.edu/CAVE/databases/columbia_gaze/
I will take a look inside these two datasets
The shape_predictor_68_face_landmarks.dat model includes pupil positions. I'm not sure how accurate it is.
As now there is a sample head position estimation and face landmark detection, I was wondering is there any plan to add gaze estimation module to Dlib library?
following repositories can be useful to do so: https://github.com/mpatacchiola/deepgaze https://github.com/opengazer/OpenGazer