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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add separate gaze and appearance encoders (ablation) | GazeShift · VRGaze | mean gaze error (deg): 2.15 -> 2.10 | Improved |
| add cross-attention based redirection (ablation) | GazeShift · VRGaze | mean gaze error (deg): 2.10 -> 2.07 | Improved |
| add gaze-focused reconstruction loss (ablation) | GazeShift · VRGaze | mean gaze error (deg): 2.07 -> 1.84 | Improved |
| sharpen gaze loss weighting, gamma 1.0 to 2.0 (ablation) | GazeShift · VRGaze | mean gaze error (deg): 1.84 -> 2.19 | Hurt |
| sharpen gaze loss weighting, gamma 1.0 to 4.0 (ablation) | GazeShift · VRGaze | mean gaze error (deg): 1.84 -> 2.41 | Hurt |
| soften gaze loss weighting, gamma 1.0 to 0.5 (ablation) | GazeShift · VRGaze | mean gaze error (deg): 1.84 -> 2.03 | Hurt |
| add calibration-aware loss to supervised model (ablation) | Supervised appearance model · VRGaze | mean gaze error (deg): 1.64 -> 1.54 | Improved |
| train on on-axis OpenEDS2020 instead of VRGaze (ablation) | GazeShift · VRGaze (per-person) | mean gaze error (deg): 1.84 -> 5.2 | Hurt |
| use GazeShift instead of Cross-Encoder | unsupervised gaze learning · OpenEDS2020 person-agnostic | mean gaze error (deg): 5.20 -> 4.20 | Improved |
| use GazeShift instead of Cross-Encoder | unsupervised gaze learning · VRGaze per-person | mean gaze error (deg): 2.15 -> 1.84 | Improved |
| scale GazeShift gaze encoder from MobileNetV2 to ResNet-18 | GazeShift · MPIIGaze (Columbia-trained) | mean gaze error (deg): 8.00 -> 7.56 | Improved |
| use GazeShift instead of Cross-Encoder (unsupervised fine-tuning) | unsupervised gaze learning · MPIIGaze | mean gaze error (deg): 7.20 -> 7.15 | No clear effect |
Closest papers by meaning
- Automatic Gaze Analysis: A Survey of Deep Learning Based Approaches (TPAMI 2023 · similarity 0.89)
- UVAGaze: Unsupervised 1-to-2 Views Adaptation for Gaze Estimation (AAAI 2024 · similarity 0.88)
- Gaze-LLE: Gaze Target Estimation via Large-Scale Learned Encoders (CVPR 2025 · similarity 0.87)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (CVPR 2025 · similarity 0.87)
- OmniGaze: Reward-inspired Generalizable Gaze Estimation in the Wild (NeurIPS 2025 · similarity 0.86)
- Foveated Instance Segmentation (CVPR 2025 · similarity 0.86)
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.
