Research map / 3D vision / Human pose estimation
Towards High Performance Low Complexity Calibration in Appearance Based Gaze Estimation
Zhaokang Chen, Bertram Emil Shi
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 |
|---|---|---|---|
| train with gaze decomposition (ablation) | Baseline CNN · MPIIGaze (no calibration) | mean angular error: 5.3 -> 4.7 | Improved |
| train with gaze decomposition (ablation) | Dilated CNN · MPIIGaze (no calibration) | mean angular error: 4.8 -> 4.5 | Improved |
| train with gaze decomposition (ablation) | ResNet · MPIIGaze (no calibration) | mean angular error: 5.4 -> 5.2 | Improved |
| train with gaze decomposition (ablation) | Dilated ResNet · MPIIGaze (no calibration) | mean angular error: 5.3 -> 4.9 | Improved |
| replace max-pooling with dilated convolutions (ablation) | Baseline CNN + GD · MPIIGaze (no calibration) | mean angular error: 4.7 -> 4.5 | Improved |
| replace max-pooling with dilated convolutions (ablation) | ResNet · MPIIGaze (no calibration) | mean angular error: 5.4 -> 5.3 | Improved |
| replace max-pooling with dilated convolutions (ablation) | GEDDNet · cross-dataset NISLGaze to MPIIGaze | mean angular error: 7.6 -> 7.2 | Improved |
| use MAP bias estimate instead of ML estimate (ablation) | GEDDNet · MPIIGaze MGTC (T=1) | mean angular error: 3.7 -> 3.5 | Improved |
| train with gaze decomposition before calibration (ablation) | Baseline CNN · MPIIGaze SGTC (S=16) | mean angular error: 3.9 -> 3.1 | Improved |
| increase calibration targets T from 1 to 128 (GEDDNet) | GEDDNet · MPIIGaze MGTC | mean angular error: 3.5 -> 2.6 | Improved |
| use more complex calibration (LA) with many targets | LA vs GEDDNet · MPIIGaze MGTC (T=64) | mean angular error: 2.6 -> 2.5 | No clear effect |
| increase sample period from 0.1s to 0.5s | GEDDNet SGTC · NISLGaze (8 samples) | mean angular error: 3.38 -> 3.16 | Improved |
Closest papers by meaning
- Auto-Calibrated Gaze Estimation Using Human Gaze Patterns (IJCV 2017 · similarity 0.91)
- MPIIGaze: Real-World Dataset and Deep Appearance-Based Gaze Estimation (TPAMI 2017 · similarity 0.90)
- A Differential Approach for Gaze Estimation (TPAMI 2019 · similarity 0.90)
- Enhancing Accuracy of Uncertainty Estimation in Appearance-based Gaze Tracking with Probabilistic Evaluation and Calibration (CVPR 2026 · similarity 0.90)
- Appearance-Based Gaze Estimation With Deep Learning: A Review and Benchmark (TPAMI 2024 · similarity 0.90)
- Suppressing Uncertainty in Gaze Estimation (AAAI 2024 · similarity 0.88)
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.
