Research map / 3D vision / Human pose estimation
Appearance-Based Gaze Estimation via Evaluation-Guided Asymmetric Regression
Yihua Cheng, Feng Lu, Xucong Zhang
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 |
|---|---|---|---|
| add the E-Net evaluation network to AR-Net (ablation) | AR-Net · Modified MPIIGaze (leave-one-person-out) | angular error in degrees (lower is better): 5.7 -> 5.0 | Improved |
| add the E-Net evaluation network to AR-Net (ablation) | AR-Net · MPIIGaze (trained on UT Multiview) | angular error in degrees (lower is better): 9.4 -> 8.8 | Improved |
| add the E-Net evaluation network to AR-Net (ablation) | AR-Net · EyeDiap (trained on UT Multiview) | angular error in degrees (lower is better): 15.2 -> 13.5 | Improved |
| add the E-Net evaluation network to AR-Net (ablation) | AR-Net · Modified MPIIGaze (leave-one-person-out), difference between better and worse eye | better/worse eye difference in degrees: 0.7 -> 0.4 | Improved |
| use four-stream AR-Net with asymmetric loss instead of Single Eye network | Single Eye · Modified MPIIGaze (leave-one-person-out) | angular error in degrees (lower is better): 6.3 -> 5.7 | Improved |
| use four-stream AR-Net instead of Single Eye network | Single Eye · MPIIGaze (trained on UT Multiview) | angular error in degrees (lower is better): 11.8 -> 9.4 | Improved |
| use four-stream AR-Net instead of Single Eye network | Single Eye · EyeDiap (trained on UT Multiview) | angular error in degrees (lower is better): 15.6 -> 15.2 | Improved |
| give both real eye images instead of replacing the right eye with a random eye (left eye input) (ablation) | ARE-Net · Modified MPIIGaze, one eye missing | angular error in degrees (lower is better): 8.5 -> 6.3 | Improved |
| replace missing eye with a fixed frontal eye instead of a black image (left eye input) (ablation) | ARE-Net · Modified MPIIGaze, one eye missing | angular error in degrees (lower is better): 6.3 -> 10.7 | Hurt |
| copy the input eye instead of a black image for the missing eye (left eye input) (ablation) | ARE-Net · Modified MPIIGaze, one eye missing | angular error in degrees (lower is better): 6.3 -> 6.1 | No clear effect |
Closest papers by meaning
- A Differential Approach for Gaze Estimation (TPAMI 2019 · similarity 0.88)
- A Coarse-to-Fine Adaptive Network for Appearance-Based Gaze Estimation (AAAI 2020 · similarity 0.88)
- Generalizing Gaze Estimation With Outlier-Guided Collaborative Adaptation (ICCV 2021 · similarity 0.87)
- MoEGaze: A Mixture of Experts Approach for Generalizable Gaze Estimation (AAAI 2026 · similarity 0.87)
- MPIIGaze: Real-World Dataset and Deep Appearance-Based Gaze Estimation (TPAMI 2017 · similarity 0.87)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (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.
