Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Enhancing Accuracy of Uncertainty Estimation in Appearance-based Gaze Tracking with Probabilistic Evaluation and Calibration
Qiaojie Zheng, Jiucai Zhang, Amy Zhang, Xiaoli Zhang
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 post-hoc isotonic calibration | heteroskedastic gaze model ResNet18 · MPIIGaze cross-subject | Avg CPE (%): 23.17 -> 5.18 | Improved |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet18 · RTGene cross-subject | Avg CPE (%): 19.60 -> 5.26 | Improved |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet50 · RTGene to MPIIGaze cross-dataset | Avg CPE (%): 26.36 -> 4.79 | Improved |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet18 · MPIIGaze to RTGene cross-dataset | Avg angular error (deg): 13.71 -> 10.12 | Improved |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet18 · MPIIGaze cross-subject | Avg angular error (deg): 5.77 -> 5.09 | Improved |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet50 · RTGene to MPIIGaze cross-dataset | Avg angular error (deg): 13.76 -> 14.24 | Hurt |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet18 · MPIIGaze cross-subject | 95% CI coverage (%): 41.1 -> 88.0 | Improved |
| add post-hoc isotonic calibration | heteroskedastic gaze model ResNet50 · RTGene to MPIIGaze cross-dataset | 95% CI coverage (%): 46.2 -> 88.6 | Improved |
| use calibration then measure with EUC | heteroskedastic gaze model ResNet18 · MPIIGaze cross-subject | Avg EUC: 0.14 -> 0.26 | Mixed |
| use calibration then measure with EUC | heteroskedastic gaze model ResNet50 · RTGene to MPIIGaze cross-dataset | Avg EUC: -0.04 -> 0.14 | Improved |
| increase calibration samples from 10 to 50-100 (ablation) | isotonic calibration · RTGene cross-subject test fold | CPE: 40.1% -> near zero | Improved |
| use calibrated quantiles vs quantile regression for 95% CI | quantile regression · MPIIGaze cross-subject | 95% CI coverage (%): 40.5 -> 88.0 | Improved |
Closest papers by meaning
- Towards High Performance Low Complexity Calibration in Appearance Based Gaze Estimation (TPAMI 2022 · similarity 0.90)
- Suppressing Uncertainty in Gaze Estimation (AAAI 2024 · similarity 0.88)
- Generalizing Eye Tracking With Bayesian Adversarial Learning (CVPR 2019 · similarity 0.86)
- Multi-scale Object-Aware Gaze Estimation via Geometric Reasoning (ECCV 2026 · similarity 0.85)
- A Generalized Label Shift Perspective for Cross-Domain Gaze Estimation (NeurIPS 2025 · similarity 0.84)
- ST-DiffEye: Diffusion-based Continuous Gaze Generation via Joint Scanpath-Trajectory Modeling (NeurIPS 2026 · similarity 0.83)
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.
