Research map / 3D vision / Face and shape
Instant Expressive Gaussian Head Avatars at Over 100 FPS
Kaiwen Jiang, Xueting Li, Seonwook Park, Ravi Ramamoorthi, Shalini De Mello, Koki Nagano
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 |
|---|---|---|---|
| deform Gaussians in feature space instead of spatial attributes (ablation) | Instant Gaussian avatar · Self-reenactment ablation | AED: 0.634 -> 0.507 | Improved |
| feature-space instead of spatial deformation, speed cost (ablation) | Instant Gaussian avatar · Self-reenactment ablation | FPS: 141.61 -> 132.87 | Hurt |
| use lightweight X-NeMo motion encoder instead of DINO-v2 (ablation) | Instant Gaussian avatar · Self-reenactment ablation | FPS: 22.47 -> 132.87 | Improved |
| use X-NeMo motion encoder instead of DINO-v2 (ablation) | Instant Gaussian avatar · Self-reenactment ablation | AED: 0.597 -> 0.507 | Improved |
| train on synthetic diffusion-generated expressions instead of real dataset (ablation) | Instant Gaussian avatar · Self-reenactment ablation | AED: 0.543 -> 0.507 | Improved |
| train on synthetic diffusion-generated expressions instead of real dataset (ablation) | Instant Gaussian avatar · Self-reenactment ablation | PSNR/MEt3R: PSNR 23.10 -> 23.18; MEt3R 0.073 -> 0.073 | No clear effect |
| retrain Portrait4D-v2 and GAGAvatar on the synthetic dataset | Portrait4D-v2 · Cross-reenactment | AED: 0.886 -> 0.972 | Hurt |
| use Ours instead of X-NeMo (2D diffusion teacher) | Instant Gaussian avatar · Cross-reenactment | AED: 0.805 -> 0.731 | Improved |
| use Ours instead of InvertAvatar for image fidelity | Instant Gaussian avatar · Self-reenactment | PSNR: 23.43 -> 22.34 | Hurt |
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.
- Portrait4D-v2 — Portrait4D: Learning One-Shot 4D Head Avatar Synthesis using Synthetic Data (CVPR 2024) · 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
- FlashAvatar: High-fidelity Head Avatar with Efficient Gaussian Embedding (CVPR 2024 · similarity 0.89)
- Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians (CVPR 2024 · similarity 0.88)
- RealPortrait: Realistic Portrait Animation with Diffusion Transformers (AAAI 2025 · similarity 0.88)
- GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians (CVPR 2024 · similarity 0.88)
- 3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations (CVPR 2025 · similarity 0.88)
- High-Fidelity 3D Head Avatars Reconstruction through Spatially-Varying Expression Conditioned Neural Radiance Field (AAAI 2024 · similarity 0.88)
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.
