Research map / 3D vision / Depth and stereo
GRAPE: Generalizable and Robust Multi-view Facial Capture
Jing Li, Di Kang, Zhenyu He
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 learned head localizer with visual hull (cross-camera-array) (ablation) | TEMPEH global · FaMoS -> FaceScape, no finetuning | average error (mm): 32.27 -> 8.06 | Improved |
| add visibility-aware 3D feature aggregation (ablation) | TEMPEH + V.H. init. · FaMoS -> FaceScape, no finetuning | average error (mm): 8.06 -> 7.20 | Improved |
| add update-by-disagreement training (ablation) | TEMPEH + V.H. init. + vis. · FaMoS -> FaceScape, no finetuning | average error (mm): 7.20 -> 2.43 | Improved |
| add update-by-disagreement training on noisy FaceScape (ablation) | TEMPEH + V.H. init. + vis. · FaceScape same-array | average error (mm): 2.54 -> 1.20 | Improved |
| replace learned localizer with visual hull on noisy FaceScape (ablation) | TEMPEH global · FaceScape same-array | average error (mm): 3.10 -> 2.53 | Improved |
| replace learned localizer with visual hull on clean FaMoS (ablation) | TEMPEH global · FaMoS complete head | average error (mm): 2.11 -> 2.38 | Hurt |
| add update-by-disagreement to local stage (GRAPE local vs TEMPEH local) | TEMPEH local · FaMoS face region | average error (mm): 0.43 -> 0.44 | No clear effect |
| vary visibility margin rho (0.1, 0.05, 0) (ablation) | GRAPE global · FaceScape | average error (mm): 1.20 / 1.25 / 1.23 | No clear effect |
| add visual hull init and visibility aggregation on FaceScape -> FaMoS (ablation) | TEMPEH global · FaceScape -> FaMoS, no finetuning | average error (mm): 32.26 -> 2.63 | Improved |
Closest papers by meaning
- Instant Multi-View Head Capture Through Learnable Registration (CVPR 2023 · similarity 0.85)
- RetinaFace: Single-Shot Multi-Level Face Localisation in the Wild (CVPR 2020 · similarity 0.84)
- Topologically Consistent Multi-View Face Inference Using Volumetric Sampling (ICCV 2021 · similarity 0.84)
- Detailed Facial Geometry Recovery from Multi-View Images by Learning an Implicit Function (AAAI 2022 · similarity 0.84)
- Deep Facial Non-Rigid Multi-View Stereo (CVPR 2020 · similarity 0.84)
- 3D Face Tracking from 2D Video through Iterative Dense UV to Image Flow (CVPR 2024 · similarity 0.82)
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.
