Research map / Architectures and efficiency / Quantization
POCA: Post-training Quantization with Temporal Alignment for Codec Avatars
Jian Meng, Yuecheng Li, Leo (Chenghui) Li, Syed Shakib Sarwar, Dilin Wang, Jae-sun Seo
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 |
|---|---|---|---|
| add view-aware visibility masking and shape-wise dynamic filtering to PTQ | Avatar texture decoder, W8A8 · Multiface avatar 2643814 front | FovVideoVDP: 5.81 -> 6.19 | Improved |
| apply POCA instead of standard PTQ at 6-bit | Avatar texture decoder, W6A6 · Multiface avatar 2643814 front | FovVideoVDP: 4.50 -> 6.06 | Improved |
| replace 8-bit QAT with POCA PTQ | Avatar decoder quantization · Multiface avatar 2643814 front, W8A8 | FovVideoVDP: 5.92 -> 6.19 | Improved |
| use PTQ with POCA instead of QAT | Avatar decoder quantization · Quantization time | quantization time: >10 hrs -> ~13 min | Improved |
| raise SWD threshold above 0.9 (ablation) | POCA · Avatar 2643814 | FovVideoVDP / SSIM: FovVideoVDP 6.213 -> 6.185 (tau 0.95), 6.121 (tau 0.975) | Hurt |
| use sparse SWD calibration (80% sparsity) on super-resolution | SRResNet W8A8 · Set5/Set14/BSD100/Urban100 x4 | PSNR: BSD100 27.618 (SR-PTQ) -> 27.620; Set5 32.207 -> 31.898 | Mixed |
| apply 8-bit standard PTQ to avatar decoder | Avatar decoder · Rendered avatar in motion | temporal noise: pixel shift +-0.44, +-0.47, +-0.25 on R, G, B | Hurt |
Later papers that compared with this method
1 later accepted paper reports POCA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ESCA: Enabling Seamless Codec Avatar Execution through Algorithm and Hardware Co-Optimization for Virtual Reality (NeurIPS 2025 · better on 2 of 2 shared results)
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
- ESCA: Enabling Seamless Codec Avatar Execution through Algorithm and Hardware Co-Optimization for Virtual Reality (NeurIPS 2025 · similarity 0.86)
- RAM-Avatar: Real-time Photo-Realistic Avatar from Monocular Videos with Full-body Control (CVPR 2024 · similarity 0.85)
- Codec Avatar Studio: Paired Human Captures for Complete, Driveable, and Generalizable Avatars (NeurIPS 2024 · similarity 0.84)
- High-Fidelity 3D Head Avatars Reconstruction through Spatially-Varying Expression Conditioned Neural Radiance Field (AAAI 2024 · similarity 0.84)
- GASP: Gaussian Avatars with Synthetic Priors (CVPR 2025 · similarity 0.83)
- MoSAR: Monocular Semi-Supervised Model for Avatar Reconstruction using Differentiable Shading (CVPR 2024 · similarity 0.83)
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.
