| Full name | SHeaP: Self-Supervised Head Geometry Predictor Learned via 2D Gaussians |
|---|---|
| Publication title | SHeaP: Self-Supervised Head Geometry Predictor Learned via 2D Gaussians |
| Publication authors | Liam Schoneveld, Zhe Chen, Davide Davoli, Jiapeng Tang, Saimon Terazawa, Ko Nishino, Matthias Nießner |
| Publication venue | ICCV 2025 |
| Publication url | https://arxiv.org/abs/2504.12292 |
| Project link | https://nlml.github.io/sheap/ |
| Code | Code available with license: CC BY-NC |
| Use other data | No |
| Last edited | 25 March, 2026 |
| Chamfer L1 [mm] | Chamfer L2 [mm] | Normals Similarity | |
|---|---|---|---|
| Posed Reconstruction | 2.083 | 1.407 | 0.8764 |
| Neutral Reconstruction | 1.866 | 1.262 | 0.8818 |
| Input Image | Reconstruction | Chamfer L1 Errors | Left view | Center view | Right view |
|---|---|---|---|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Chamfer L1 Errors | Reconstruction | Left view | Center view | Right view | |
|---|---|---|---|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|