Research map / Image generation / Autoregressive and latent generation
A Hierarchical Generative Model for Eye Image Synthesis and Eye Gaze Estimation
Kang Wang, Rui Zhao, Qiang Ji
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 HGM instead of [33] for image synthesis evaluation | Eye image synthesis · MPIIGaze | Error / degree: 7.8 -> 7.6 | Improved |
| Use HGM instead of [33] for image synthesis evaluation | Eye image synthesis · UTMultiview | Error / degree: 8.9 -> 8.8 | Improved |
| Train the proposed shape estimator with more samples than 3000 | UnityEye eye shape estimation · UnityEye, 10000 training samples | Eye shape prediction error: 1.8 -> 1.4 | Improved |
| Add 10000 synthesized image-shape pairs to real training data (ablation) | Proposed pupil detector · BioID, deye ≤ 0.05 | Pupil detection rate: 90.3% -> 92.1% | Improved |
| Add 10000 synthesized image-shape pairs to real training data (ablation) | Proposed pupil detector · BioID, deye ≤ 0.10 | Pupil detection rate: 99.3% -> 99.9% | Improved |
| Add 10000 synthesized image-shape pairs to real training data (ablation) | Proposed pupil detector · BioID, deye ≤ 0.25 | Pupil detection rate: 99.9% -> 99.9% | No clear effect |
| Use HGM instead of CNN[43] | Gaze estimation · UnityEye, within-dataset | Error / degree: 3.4 -> 4.5 | Hurt |
| Use HGM instead of CNN[43] | Gaze estimation · UnityEye, cross-dataset | Error / degree: 19.5 -> 15.9 | Improved |
| Use HGM instead of CNN[43] | Gaze estimation · UTMultiview, within-dataset | Error / degree: 6.8 -> 8.4 | Hurt |
| Use HGM instead of CNN[43] | Gaze estimation · UTMultiview, cross-dataset | Error / degree: 18.3 -> 14.2 | Improved |
| Use HGM instead of CNN[43] | Gaze estimation · MPIIGaze, within-dataset | Error / degree: 6.1 -> 7.5 | Hurt |
| Use HGM instead of CNN[43] | Gaze estimation · MPIIGaze, cross-dataset | Error / degree: 12.3 -> 7.7 | Improved |
Closest papers by meaning
- Photo-Realistic Monocular Gaze Redirection Using Generative Adversarial Networks (ICCV 2019 · similarity 0.89)
- Self-Learning Transformations for Improving Gaze and Head Redirection (NeurIPS 2020 · similarity 0.89)
- Deep Pictorial Gaze Estimation (ECCV 2018 · similarity 0.87)
- RGBD Based Gaze Estimation via Multi-Task CNN (AAAI 2019 · similarity 0.87)
- ReDirTrans: Latent-to-Latent Translation for Gaze and Head Redirection (CVPR 2023 · similarity 0.86)
- GazeGene: Large-scale Synthetic Gaze Dataset with 3D Eyeball Annotations (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.
