Research map / 3D vision / Face and shape
Pixel Codec Avatars
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the baseline decoder with PiCA's visible-pixel decoding and dense geometry | codec avatar rendering · S1, five test views | MSE: 23.03 -> 21.39; 20.78 -> 19.71; 18.13 -> 17.52; 16.32 -> 15.52; 18.97 -> 18.00 | Helped |
| replace the baseline decoder with PiCA's visible-pixel decoding and dense geometry | codec avatar rendering · S6, five test views | MSE: 6.86 -> 6.22; 6.53 -> 6.06; 5.85 -> 5.39; 5.66 -> 4.97; 5.29 -> 4.95 | Helped |
| decode only visible pixels, with work adapted to avatar distance | PiCA on Oculus Quest 2 · 18cm, 65cm, and 120cm | DSP time: 44.76 ms -> 2.16 ms | Helped |
| remove UV surface coordinates (ablation) | PiCA · S1, S2, and S6 across five views | MSE: 21.39 -> 22.16 (S1 Front); 18.31 -> 19.34 (S2 Front); 6.22 -> 6.86 (S6 Front) | Hurt |
| decode a coarse 7K-vertex mesh instead of dense geometry (ablation) | PiCA · S1, S2, and S6 across five views | MSE: 21.39 -> 21.64 (S1 Front); 18.31 -> 19.88 (S2 Front); 6.22 -> 6.54 (S6 Front) | Hurt |
| replace learned UV encoding maps with sinusoidal positional encoding (ablation) | PiCA · S1, S2, and S6 across five views | MSE: 21.39 -> 21.85 (S1 Front); 18.31 -> 18.99 (S2 Front); 6.22 -> 6.41 (S6 Front) | Hurt |
| use raw UV coordinates without positional encoding (ablation) | PiCA · S1, S2, and S6 across five views | MSE: 21.39 -> 21.45 (S1 Front); 18.31 -> 19.17 (S2 Front); 6.22 -> 6.59 (S6 Front) | Hurt |
| use only joint 2D UV encoding (ablation) | PiCA · S1, S2, and S6 across five views | MSE: 21.39 -> 21.56 (S1 Front); 18.31 -> 19.05 (S2 Front); 6.22 -> 6.28 (S6 Front); 6.06 -> 6.00 (S6 Up) | Mixed |
| use only separate 1D UV encodings (ablation) | PiCA · S1, S2, and S6 across five views | MSE: 21.39 -> 21.40 (S1 Front); 18.31 -> 19.30 (S2 Front); 6.22 -> 6.58 (S6 Front); 19.71 -> 19.67 (S1 Up) | Mixed |
Closest papers by meaning
- High-Fidelity 3D Head Avatars Reconstruction through Spatially-Varying Expression Conditioned Neural Radiance Field (AAAI 2024 · similarity 0.87)
- Stratified Avatar Generation from Sparse Observations (CVPR 2024 · similarity 0.87)
- ESCA: Enabling Seamless Codec Avatar Execution through Algorithm and Hardware Co-Optimization for Virtual Reality (NeurIPS 2025 · similarity 0.87)
- Codec Avatar Studio: Paired Human Captures for Complete, Driveable, and Generalizable Avatars (NeurIPS 2024 · similarity 0.87)
- Arc2Avatar: Generating Expressive 3D Avatars from a Single Image via ID Guidance (CVPR 2025 · similarity 0.86)
- NECA: Neural Customizable Human Avatar (CVPR 2024 · similarity 0.86)
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.