TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Fast Bi-layer Neural Synthesis of One-Shot Realistic Head ...

Fast Bi-layer Neural Synthesis of One-Shot Realistic Head Avatars

Egor Zakharov, Aleksei Ivakhnenko, Aliaksandra Shysheya, Victor Lempitsky

2020-08-24ECCV 2020 8Neural RenderingTalking Head Generation
PaperPDFCode(official)

Abstract

We propose a neural rendering-based system that creates head avatars from a single photograph. Our approach models a person's appearance by decomposing it into two layers. The first layer is a pose-dependent coarse image that is synthesized by a small neural network. The second layer is defined by a pose-independent texture image that contains high-frequency details. The texture image is generated offline, warped and added to the coarse image to ensure a high effective resolution of synthesized head views. We compare our system to analogous state-of-the-art systems in terms of visual quality and speed. The experiments show significant inference speedup over previous neural head avatar models for a given visual quality. We also report on a real-time smartphone-based implementation of our system.

Results

TaskDatasetMetricValueModel
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
Facial Recognition and ModellingVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
Image GenerationVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
Image GenerationVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
Image GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
Image GenerationVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
Image GenerationVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
Image GenerationVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
Image GenerationVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
Image GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
Image GenerationVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
Image GenerationVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
Image GenerationVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
Image GenerationVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
Image GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
Image GenerationVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
Image GenerationVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
Talking Head GenerationVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
Face GenerationVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
Face GenerationVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
Face GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
Face GenerationVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
Face GenerationVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
Face GenerationVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
Face GenerationVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
Face GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
Face GenerationVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
Face GenerationVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
Face GenerationVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
Face GenerationVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
Face GenerationVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
Face GenerationVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
Face GenerationVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
Face ReconstructionVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
3DVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
3DVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
3DVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
3DVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
3DVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
3DVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
3DVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
3DVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
3DVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
3DVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
3DVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
3DVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
3DVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
3DVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
3DVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
3D Face ModellingVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
3D Face ReconstructionVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningCSIM0.653Fast Bi-layer Avatars (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningLPIPS0.358Fast Bi-layer Avatars (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningNormalized Pose Error43.3Fast Bi-layer Avatars (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningSSIM0.508Fast Bi-layer Avatars (medium size)
10-shot image generationVoxCeleb2 - 1-shot learninginference time (ms)4Fast Bi-layer Avatars (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningCSIM0.638First Order Motion Model (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningLPIPS0.311First Order Motion Model (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningNormalized Pose Error47.8First Order Motion Model (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningSSIM0.553First Order Motion Model (medium size)
10-shot image generationVoxCeleb2 - 1-shot learninginference time (ms)13First Order Motion Model (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningCSIM0.604Few-shot Vid-to-vid (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningLPIPS0.368Few-shot Vid-to-vid (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningNormalized Pose Error46.1Few-shot Vid-to-vid (medium size)
10-shot image generationVoxCeleb2 - 1-shot learningSSIM0.419Few-shot Vid-to-vid (medium size)
10-shot image generationVoxCeleb2 - 1-shot learninginference time (ms)22Few-shot Vid-to-vid (medium size)

Related Papers

MEDTalk: Multimodal Controlled 3D Facial Animation with Dynamic Emotions by Disentangled Embedding2025-07-08HiNeuS: High-fidelity Neural Surface Mitigating Low-texture and Reflective Ambiguity2025-06-30Advancing Talking Head Generation: A Comprehensive Survey of Multi-Modal Methodologies, Datasets, Evaluation Metrics, and Loss Functions2025-06-23R3eVision: A Survey on Robust Rendering, Restoration, and Enhancement for 3D Low-Level Vision2025-06-19Audio-Visual Driven Compression for Low-Bitrate Talking Head Videos2025-06-16Gaussian Herding across Pens: An Optimal Transport Perspective on Global Gaussian Reduction for 3DGS2025-06-11R3D2: Realistic 3D Asset Insertion via Diffusion for Autonomous Driving Simulation2025-06-09Unifying Appearance Codes and Bilateral Grids for Driving Scene Gaussian Splatting2025-06-05