Research map / Label-efficient and robust learning / Domain adaptation and generalization
Learning a Generalized Gaze Estimator from Gaze-Consistent Feature
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| disturb identity, expression, illumination and tone, then align original and disturbed gaze features | ResNet-18 gaze estimator · ETH-XGaze → MPIIGaze | angular error (°): 8.80 -> 6.50 | Helped |
| disturb identity, expression, illumination and tone, then align original and disturbed gaze features | ResNet-18 gaze estimator · ETH-XGaze → EyeDiap | angular error (°): 8.52 -> 7.44 | Helped |
| disturb identity, expression, illumination and tone, then align original and disturbed gaze features | ResNet-18 gaze estimator · Gaze360 → MPIIGaze | angular error (°): 8.18 -> 7.55 | Helped |
| disturb identity, expression, illumination and tone, then align original and disturbed gaze features | ResNet-18 gaze estimator · Gaze360 → EyeDiap | angular error (°): 9.00 -> 9.03 | No clear effect |
| add identity disturbance (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.63 -> 8.31 | Helped |
| add expression disturbance (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.63 -> 7.65 | Helped |
| add illumination disturbance (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.63 -> 7.52 | Helped |
| add tone disturbance (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.63 -> 8.25 | Helped |
| add identity and tone disturbance (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.63 -> 7.48 | Helped |
| disturb all four gaze-irrelevant factors instead of no factors (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.63 -> 7.63 | Helped |
| set adversarial noise intensity to ϵ = 0.2 instead of ϵ = 0.1 (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 7.63 -> 9.73 | Hurt |
| align features with MMD rather than training on disturbed data without feature alignment (ablation) | gaze estimator · four domain generalization tasks (Avg) | angular error (°): 8.30 -> 7.63 | Helped |
Closest papers by meaning
- Gaze from Origin: Learning for Generalized Gaze Estimation by Embedding the Gaze Frontalization Process (AAAI 2024 · similarity 0.91)
- From Feature to Gaze: A Generalizable Replacement of Linear Layer for Gaze Estimation (CVPR 2024 · similarity 0.91)
- Gaze Label Alignment: Alleviating Domain Shift for Gaze Estimation (AAAI 2025 · similarity 0.89)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (CVPR 2025 · similarity 0.87)
- CLIP-Gaze: Towards General Gaze Estimation via Visual-Linguistic Model (AAAI 2024 · similarity 0.86)
- A Generalized Label Shift Perspective for Cross-Domain Gaze Estimation (NeurIPS 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.