Research map / Label-efficient and robust learning / Recognition and long-tail
A Coarse-to-Fine Adaptive Network for Appearance-Based Gaze Estimation
Yihua Cheng, Shiyao Huang, Fei Wang, Chen Qian, Feng Lu
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 |
|---|---|---|---|
| estimate a basic gaze direction from face features, then refine it with eye-derived residuals (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error: 4.65^{\circ} -> 4.14^{\circ} | Improved |
| add a learned gate to connect face-based basic gaze and eye-based residual prediction (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error: 4.14^{\circ} -> 4.46^{\circ} | Improved |
| add adaptive attention for weighting left and right eye features (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error: 4.14^{\circ} -> 4.50^{\circ} | Improved |
| use coarse-to-fine rather than fine-to-coarse estimation (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error, refined: 5.00^{\circ} -> 4.14^{\circ} | Improved |
| condition residual prediction on the basic gaze state using a bi-gram connection (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error, refined: 4.43^{\circ} -> 4.14^{\circ} | Improved |
| use both face and corresponding eye features to generate eye-attention weights (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error, refined: 4.30^{\circ} -> 4.14^{\circ}; 4.34^{\circ} -> 4.14^{\circ} | Improved |
| combine coarse-to-fine face and eye processing instead of using face-only estimation (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error: 4.58^{\circ} -> 4.14^{\circ} | Improved |
| combine coarse-to-fine processing instead of concatenating face and eye features (ablation) | CA-Net · MPIIGaze evaluation set | mean angular error: 5.00^{\circ} -> 4.14^{\circ} | Improved |
| use CA-Net rather than RT-Gene (4 ensemble) | gaze estimation · MPIIGaze evaluation set | mean angular error: 4.3^{\circ} -> 4.1^{\circ} | Improved |
| use CA-Net rather than Dilated-Net | gaze estimation · EyeDiap screen target | mean angular error: 5.9^{\circ} -> 5.3^{\circ} | Improved |
Closest papers by meaning
- A Differential Approach for Gaze Estimation (TPAMI 2019 · similarity 0.89)
- Generalizing Gaze Estimation With Outlier-Guided Collaborative Adaptation (ICCV 2021 · similarity 0.88)
- Appearance-Based Gaze Estimation via Evaluation-Guided Asymmetric Regression (ECCV 2018 · similarity 0.88)
- MPIIGaze: Real-World Dataset and Deep Appearance-Based Gaze Estimation (TPAMI 2017 · similarity 0.88)
- UVAGaze: Unsupervised 1-to-2 Views Adaptation for Gaze Estimation (AAAI 2024 · similarity 0.86)
- FIFA: Fine-grained Inter-frame Attention for Driver's Video 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.
