Gaussian Pixel Codec Avatars: A Hybrid Representation for Efficient Rendering
Not on the map. Why?
Abstract (arXiv)
We present Gaussian Pixel Codec Avatars (GPiCA), photorealistic head avatars that can be generated from multi-view images and efficiently rendered on mobile devices. GPiCA utilizes a unique hybrid representation that combines a triangle mesh and anisotropic 3D Gaussians. This combination maximizes memory and rendering efficiency while maintaining a photorealistic appearance. The triangle mesh is highly efficient in representing surface areas like facial skin, while the 3D Gaussians effectively handle non-surface areas such as hair and beard. To this end, we develop a unified differentiable rendering pipeline that treats the mesh as a semi-transparent layer within the volumetric rendering paradigm of 3D Gaussian Splatting. We train neural networks to decode a facial expression code into three components: a 3D face mesh, an RGBA texture, and a set of 3D Gaussians. These components are rendered simultaneously in a unified rendering engine. The networks are trained using multi-view image supervision. Our results demonstrate that GPiCA achieves the realism of purely Gaussian-based avatars while matching the rendering performance of mesh-based avatars.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 34 of its 47 references that are in the map. First posted on arXiv 17 Dec 2025.
Builds on
34 of 47 references.
References · 34
- 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
- 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
- 2022Generative Adversarial Networkscited by 5,627
- 2021Efficient Geometry-aware 3D Generative Adversarial Networkscited by 1,205
- 20242D Gaussian Splatting for Geometrically Accurate Radiance Fieldscited by 1,076
- 2023SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Renderingcited by 616
- 2023LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPScited by 436
- 2023Compact 3D Gaussian Representation for Radiance Fieldcited by 372
- 2021Mixture of Volumetric Primitives for Efficient Neural Renderingcited by 277
- 20193D Morphable Face Models -- Past, Present and Futurecited by 270
- 2021I M Avatar: Implicit Morphable Head Avatars from Videoscited by 231
- 2021Neural Head Avatars from Monocular RGB Videoscited by 213
22 more
- 2023NeRSemble: Multi-view Radiance Field Reconstruction of Human Headscited by 194
- 2024SplattingAvatar: Realistic Real-Time Human Avatars with Mesh-Embedded Gaussian Splattingcited by 172
- 2018Deep Appearance Models for Face Renderingcited by 164
- 2024High-quality Surface Reconstruction using Gaussian Surfelscited by 161
- 2022Instant Volumetric Head Avatarscited by 159
- 2023Compact 3D Scene Representation via Self-Organizing Gaussian Gridscited by 151
- 2017Morphable Face Models - An Open Frameworkcited by 138
- 2023NeuSG: Neural Implicit Surface Reconstruction with 3D Gaussian Splatting Guidancecited by 138
- 2021Pixel Codec Avatarscited by 127
- 2023Relightable Gaussian Codec Avatarscited by 127
- 2022Next3D: Generative Neural Texture Rasterization for 3D-Aware Head Avatarscited by 124
- 2022Realistic One-shot Mesh-based Head Avatarscited by 119
- 2023FlashAvatar: High-fidelity Head Avatar with Efficient Gaussian Embeddingcited by 108
- 2024GES: Generalized Exponential Splatting for Efficient Radiance Field Renderingcited by 92
- 2022Multiface: A Dataset for Neural Face Renderingcited by 87
- 2024GaMeS: Mesh-Based Adapting and Modification of Gaussian Splattingcited by 70
- 2024MeGA: Hybrid Mesh-Gaussian Head Avatar for High-Fidelity Rendering and Head Editingcited by 38
- 2024HAHA: Highly Articulated Gaussian Human Avatars with Textured Mesh Priorcited by 31
- 2024Gaussian Frosting: Editable Complex Radiance Fields with Real-Time Renderingcited by 19
- 2023GAN-Avatar: Controllable Personalized GAN-based Human Head Avatarcited by 14
- 2024Rasterized Edge Gradients: Handling Discontinuities Differentiablycited by 14
- 2024Direct and Adaptable Mesh-Gaussian Scene Reconstruction from Multi-View Imagescited by 9
Built on by
Not cited yet.