List view
Add the April tag detection pipeline and run it over an entire session and compare its performance i.e. accuracy and run time.
Overdue by 4 year(s)•Due by November 5, 2021The intrinsics and extrinsics calculated by the pupil pipeline are inaccurate especially for the FLIR (or wide-angle cameras) This shows itself in 2D gaze estimation the most. Therefore in this milestone, the initial gaze calibration/validation portion of a recorded session is used to estimate both intrinsics and extrinsics.
Overdue by 4 year(s)•Due by November 5, 2021Add 3D gaze calibration pipeline to the repo. So far the monocular and binocular 2D calibrations are implemented. The 3D calibration is yet to be merged and tested properly. Which does depend on correct intrinsics and extrinsincs matrices.
Overdue by 4 year(s)•Due by November 12, 2021•1/1 issues closedCurrently, the only pupil tracking pipeline implemented is pupil labs. DeepVog performs mediocre only on the indoor data. This milestone is to add the DeepVog implementation, make sure their 3D monocular gaze works fine with the correct eye camera intrinsics and compare its accuracy with the other methods.
Overdue by 4 year(s)•Due by November 12, 2021Currently, the only pupil tracking pipeline implemented is pupil labs. RITnet performs mediocre only on the indoor data. This milestone is to bring the RITnet implementation from my RGnet repo, retrain the network and add the RITnet pipeline to the repo.
Overdue by 4 year(s)•Due by November 12, 2021Different analysis results are saved into different formats currently. The major issue is inconsistency and the time that it takes to be loaded afterward. Therefore, this milestone is in relation to the data-saving one.
Overdue by 4 year(s)•Due by November 5, 2021•1/1 issues closed