PTGE

Personal Transformer-Based Gaze Estimation

Implementation of the paper http://jcse.kiise.org/files/V17N2-01.pdf

Requirements

tensorflow==2.15.1 tf-models-official==2.15

In case the notebooks are not rendered try them here https://nbviewer.org/

Download MPIIGaze dataset http://datasets.d2.mpi-inf.mpg.de/MPIIGaze/MPIIGaze.tar.gz

Download MPIIFaceGaze dataset http://datasets.d2.mpi-inf.mpg.de/MPIIGaze/MPIIFaceGaze.zip

For preparing the data

$python prep_data.py

Note: Edit the MPII dataset paths and the number of persons in prep_data.py

prep_data.py.-.PTGE.WSL_.Ubuntu-22.04.-.Visual.Studio.Code.2024-05-21.20-06-42.mp4

Run the train_GazeModel.ipynb for implementation and training of the GazeModel

train_GazeModel._.Kaggle.-.Google.Chrome.2024-05-22.12-08-44.mp4

Run the train_CalibrationModel.ipynb for implementation and training of the CalibrationModel and calibrate it for new user

Run the test_calibration.ipynb for using newly estimated person embeddings from calibrationModel in the gazemodel