AI Paper Atlas

Research map / 3D vision / Human pose estimation

GazeShift: Unsupervised Gaze Estimation and Dataset for VR

Gil Shapira, Ishay Goldin, Evgeny Artyomov, Donghoon Kim, Yosi Keller, Niv Zehngut

CVPR 2026 · 3D vision › Human pose estimation · approach: Learned camera pose estimation · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add separate gaze and appearance encoders (ablation)GazeShift · VRGazemean gaze error (deg): 2.15 -> 2.10Improved
add cross-attention based redirection (ablation)GazeShift · VRGazemean gaze error (deg): 2.10 -> 2.07Improved
add gaze-focused reconstruction loss (ablation)GazeShift · VRGazemean gaze error (deg): 2.07 -> 1.84Improved
sharpen gaze loss weighting, gamma 1.0 to 2.0 (ablation)GazeShift · VRGazemean gaze error (deg): 1.84 -> 2.19Hurt
sharpen gaze loss weighting, gamma 1.0 to 4.0 (ablation)GazeShift · VRGazemean gaze error (deg): 1.84 -> 2.41Hurt
soften gaze loss weighting, gamma 1.0 to 0.5 (ablation)GazeShift · VRGazemean gaze error (deg): 1.84 -> 2.03Hurt
add calibration-aware loss to supervised model (ablation)Supervised appearance model · VRGazemean gaze error (deg): 1.64 -> 1.54Improved
train on on-axis OpenEDS2020 instead of VRGaze (ablation)GazeShift · VRGaze (per-person)mean gaze error (deg): 1.84 -> 5.2Hurt
use GazeShift instead of Cross-Encoderunsupervised gaze learning · OpenEDS2020 person-agnosticmean gaze error (deg): 5.20 -> 4.20Improved
use GazeShift instead of Cross-Encoderunsupervised gaze learning · VRGaze per-personmean gaze error (deg): 2.15 -> 1.84Improved
scale GazeShift gaze encoder from MobileNetV2 to ResNet-18GazeShift · MPIIGaze (Columbia-trained)mean gaze error (deg): 8.00 -> 7.56Improved
use GazeShift instead of Cross-Encoder (unsupervised fine-tuning)unsupervised gaze learning · MPIIGazemean gaze error (deg): 7.20 -> 7.15No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy