Research map / 3D vision / NeRF and radiance fields
GazeNeRF: 3D-Aware Gaze Redirection With Neural Radiance Fields
Alessandro Ruzzi, Xiangwei Shi, Xi Wang, Gengyan Li, Shalini De Mello, Hyung Jin Chang, Xucong Zhang, Otmar Hilliges
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 |
|---|---|---|---|
| Use separate face and eye streams instead of a single MLP (ablation) | GazeNeRF · ETH-XGaze | Gaze error: 11.427 -> 8.609 | Improved |
| Add rotation to the eye stream (ablation) | Two-stream GazeNeRF · ETH-XGaze | Gaze error: 8.609 -> 8.437 | Improved |
| Add a rotation matrix to the single-stream feature maps (ablation) | vanilla-GazeNeRF · ETH-XGaze | Gaze error: 11.427 -> 9.279 | Improved |
| Add functional loss (ablation) | vanilla-GazeNeRF · ETH-XGaze | Gaze error: 11.427 -> 7.777 | Improved |
| Add functional loss to the two-stream, rotated model (ablation) | GazeNeRF · ETH-XGaze | Gaze error: 8.437 -> 6.944 | Improved |
| Add perceptual loss to reconstruction-only training (ablation) | Two-stream+rotation GazeNeRF · ETH-XGaze | Gaze error: 28.122 -> 8.861 | Improved |
| Add disentanglement loss after reconstruction and perceptual losses (ablation) | Two-stream+rotation GazeNeRF · ETH-XGaze | Head Pose: 3.456 -> 3.386 | Improved |
| Add functional loss after reconstruction, perceptual, and disentanglement losses (ablation) | Two-stream+rotation GazeNeRF · ETH-XGaze | Gaze error: 8.460 -> 6.944 | Improved |
| Add functional loss after reconstruction, perceptual, and disentanglement losses (ablation) | Two-stream+rotation GazeNeRF · ETH-XGaze | FID: 72.044 -> 81.816 | Hurt |
| Initialize with pre-trained HeadNeRF rather than train from scratch (ablation) | GazeNeRF · ETH-XGaze | Head Pose: 3.470 -> 3.080 | Improved |
| Initialize with pre-trained HeadNeRF rather than train from scratch (ablation) | GazeNeRF · ETH-XGaze | Gaze error: 6.944 -> 7.134 | Hurt |
| Use GazeNeRF instead of HeadNeRF | GazeNeRF · MPIIFaceGaze | Gaze error: 14.320 -> 14.933 | Hurt |
Later papers that compared with this method
1 later accepted paper reports GazeNeRF as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- RTGaze: Real-Time 3D-Aware Gaze Redirection from a Single Image (AAAI 2026 · better on 9 of 10 shared results)
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.
- HeadNeRF — HeadNeRF: A Real-Time NeRF-Based Parametric Head Model (CVPR 2022) · better on 7 of 12 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
- RTGaze: Real-Time 3D-Aware Gaze Redirection from a Single Image (AAAI 2026 · similarity 0.91)
- GazeGaussian: High-Fidelity Gaze Redirection with 3D Gaussian Splatting (ICCV 2025 · similarity 0.91)
- GA3CE: Unconstrained 3D Gaze Estimation with Gaze-Aware 3D Context Encoding (CVPR 2025 · similarity 0.87)
- De^2Gaze: Deformable and Decoupled Representation Learning for 3D Gaze Estimation (CVPR 2025 · similarity 0.87)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (CVPR 2025 · similarity 0.87)
- 3D Prior Is All You Need: Cross-Task Few-shot 2D Gaze Estimation (CVPR 2025 · 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.
