Finally, is there any accuracy/precision data available for the DIY headset?
I built it in 2014...
@user-36b0a3 yes staying about 0.6 to 0.8 is recommended. We cut the gaze datastream of at a threshold that you set in the settings, thus you dont see anything below. 0.6 in your case.
@user-36b0a3 we have not data for DIY headsets. Sorry.
@mpk Got it. Thanks!
@mpk the confidence threshold that you mentioned to abatty - in which settings is this to be set? Pupil Capture?
Yes in pupil capture
@mpk thanks, despite looking intensively and rereading the User Guide, I didn't find where to change this parameter. Can you give me a hint? (I'm on v1.16.43 ifthat matters.)
@user-141bcd calibration plugin menu has confidence threshold slider.
thanks @wrp this slider I know, but isn't it to set the min confidence required for a data point to be considered for calibration? I observed the same as @user-36b0a3 described (imo), namely when streaming confidence values into Unity (in my case) I would never see values <0.6. This also does not change when playing with the slider in the calibration plugin. However, I realized that this is only the case for data in the binocular mapping context (monocular data points also have smaller conf. values). I assume therefore that the Binoc. Gaze Mapper has an integrated threshold of some kind which cannot simply be changed via GUI.
It's been a while since I've played around with 3D calibration. Is it still based on grids? I ask because I came across a reference that suggests 3D calibration accuracy can be greatly improved by using a continuous-type animation of a 3D point, e.g., along a Lissajous-knot path
Duchowski, A. T. (2017). Eye Tracking Methodology: Theory and Practice (3rd ed.). Springer Publishing Company, Incorporated.
Wang, R., Pelfrey, B., Duchowski, A. T.,&House, D. H. (2012). Online gaze disparity via binocular eye tracking on stereoscopic displays. In Second Joint 3DIM/3DPVT Conference: 3D Imaging, Modeling, Processing, Visualization & Transmission (3DimPVT 2012). Zurich, Switzerland.
Hey everyone. I'm working on a research project where we use the pupil capture 120 hz add-on with a vive. I'm currently using the 3d calibration technique that is shipped with the unity sample project to calibrate the headset. Unfortunately I'm noticing that slippage causes some severe errors in the eye tracking. Has anyone looked into chin mounts that work with the vive/pupil capture system?