Research map / 3D vision / Human pose estimation
6DoF Head Pose Estimation through Explicit Bidirectional Interaction with Face Geometry
Sungho Chun, Ju Yong Chang
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 |
|---|---|---|---|
| increase bidirectional iterations from 1 to 3 (ablation) | TRG (w/o multi-scale) · ARKitFace test | ADD (mm): 8.93 -> 8.71 | Improved |
| increase bidirectional iterations from 1 to 3 (ablation) | TRG (w/o multi-scale) · BIWI | ADD (mm): 32.28 -> 31.52 | Improved |
| add multi-scale features to the 3-iteration model (ablation) | TRG · BIWI | ADD (mm): 31.52 -> 29.46 | Improved |
| estimate bounding box correction parameters instead of directly predicting translation (ablation) | TRG · BIWI (out-of-distribution) | ADD (mm): 5.1K -> 29.46 | Improved |
| estimate correction parameters instead of translation on in-distribution data (ablation) | TRG · ARKitFace test | ADD (mm): 11.66 -> 8.68 | Improved |
| add facial geometry estimation (grid sampled baseline vs landmark-free) (ablation) | TRG · ARKitFace test | ADD (mm): 9.34 -> 8.99 | Improved |
| add landmark-to-image alignment sampling (vs grid sampling) (ablation) | TRG · BIWI | ADD (mm): 35.04 -> 29.46 | Improved |
| add landmark-to-image alignment sampling on ARKitFace (ablation) | TRG · ARKitFace test | ADD (mm): 8.99 -> 8.68 | Improved |
| train additionally on 300W-LP | TRG · BIWI | ADD (mm): 32.10 -> 29.46 | Improved |
| TRG versus PerspNet on BIWI translation x-axis error | TRG · BIWI | tx error (mm): 8.41 vs 4.15 (PerspNet) | Hurt |
| increase iterations 2 to 3 on ARKitFace translation error (ablation) | TRG (w/o multi-scale) · ARKitFace test | MAEt (mm): 3.61 -> 3.63 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- img2pose — img2pose: Face Alignment and Detection via 6DoF, Face Pose Estimation (CVPR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- img2pose: Face Alignment and Detection via 6DoF, Face Pose Estimation (CVPR 2021 · similarity 0.88)
- HOGWARTS: Mitigating Perspective Distortion for 6DoF Head Pose Estimation via Virtual Camera Space (NeurIPS 2026 · similarity 0.88)
- FAR: Flexible Accurate and Robust 6DoF Relative Camera Pose Estimation (CVPR 2024 · similarity 0.86)
- TokenHPE: Learning Orientation Tokens for Efficient Head Pose Estimation via Transformers (CVPR 2023 · similarity 0.86)
- Semi-Supervised Unconstrained Head Pose Estimation in the Wild (TPAMI 2025 · similarity 0.86)
- LUVLi Face Alignment: Estimating Landmarks' Location, Uncertainty, and Visibility Likelihood (CVPR 2020 · 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.
