Research map / Learning theory and algorithms / Prediction and decision losses
Coordinate Singularities Break Conformal Coverage for Gaze and Head Pose
Mohammadreza Jamalifard, Yaxiong Lei, Parastoo Azizinezhad, Javier Andreu-Perez
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 |
|---|---|---|---|
| replace yaw-pitch L2 conformal score with geodesic score (ablation) | Split conformal gaze prediction (ResNet-18) · ETH-XGaze near pole | slice coverage (%): 38.9 -> 74.6 | Improved |
| replace yaw-pitch L2 score with geodesic score (ablation) | Split conformal gaze prediction (ResNet-18) · Gaze360 near pole | slice coverage (%): 42.0 -> 77.2 | Improved |
| add scalar normalisation to yaw-pitch score (ablation) | YP L2 conformal score · ETH-XGaze near pole | slice coverage (%): 38.9 -> 50.4 | Improved |
| add normalisation on top of geodesic score (ablation) | Geodesic conformal score · ETH-XGaze near pole | slice coverage (%): 74.6 -> 89.3 | Improved |
| use first-order metric-corrected score instead of YP L2 (ablation) | YP L2 conformal score · ETH-XGaze near pole | slice coverage (%): 38.9 -> 74.1 | Improved |
| replace Euler L2 with geodesic score (ablation) | Controlled SO(3) conformal (BIWI, sigma 5deg) · BIWI |beta|>60 | slice coverage (%): 57.5 -> 89.5 | Improved |
| replace Euler L2 with geodesic score (ablation) | SixDRepNet conformal head pose · AFLW2000-3D |beta|>60 | slice coverage (%): 55.2 -> 75.1 | Improved |
| add normalisation to Euler score (ablation) | Euler L2 conformal score · BIWI |beta|>60 | slice coverage (%): 57.5 -> 72.7 | Improved |
| add normalisation to Euler score (ablation) | Euler L2 conformal score · AFLW2000-3D |beta|>60 | slice coverage (%): 55.2 -> 62.8 | Improved |
| add Mondrian 3-bin calibration to geodesic score (ablation) | Geodesic conformal score · AFLW2000-3D |beta|>60 | slice coverage (%): 75.1 -> 90.1 | Improved |
| use stronger ViT backbone (GazeTR) with YP L2 score (ablation) | YP L2 conformal score · ETH-XGaze near pole | slice coverage (%): 38.9 -> 30.3 | Hurt |
| replace multi-chart atlas with geodesic score (ablation) | Gaze360 conformal · Gaze360 near pole | slice coverage (%): 59.7 -> 77.2 | Improved |
Closest papers by meaning
- Task-Induced Riemannian Metrics for Vision Transformer Feature Spaces (NeurIPS 2026 · similarity 0.85)
- 3DGazeNet: Generalizing Gaze Estimation with Weak Supervision from Synthetic Views (ECCV 2024 · similarity 0.85)
- MAtCha Gaussians: Atlas of Charts for High-Quality Geometry and Photorealism From Sparse Views (CVPR 2025 · similarity 0.85)
- FoundationGeo: Learning Spatial Pixel-Wise Fields for Monocular Metric Geometry (ECCV 2026 · similarity 0.85)
- Geometric Collapse: When Vision Models Fail to Verify Physical Causality (ICML 2026 · similarity 0.85)
- Generalizing Visual Geometry Priors to Sparse Gaussian Occupancy Prediction (CVPR 2026 · similarity 0.85)
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.
