Research map / 3D vision / Human pose estimation
GazeOnce360: Fisheye-Based 360deg Multi-Person Gaze Estimation with Global-Local Feature Fusion
Zhuojiang Cai, Zhenghui Sun, 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 |
|---|---|---|---|
| add rotational convolution (ablation) | GazeOnce360 · MPSGaze360 test split | gaze error (deg): 12.14 -> 11.14 | Improved |
| add eye landmark supervision on top of RotConv (ablation) | GazeOnce360 · MPSGaze360 test split | gaze error (deg): 11.14 -> 8.890 | Improved |
| add face landmark supervision on top of RotConv (ablation) | GazeOnce360 · MPSGaze360 test split | gaze error (deg): 11.14 -> 9.782 | Improved |
| add face landmarks in addition to eye landmarks (ablation) | GazeOnce360 · MPSGaze360 test split | gaze error (deg): 8.890 -> 8.945 | No clear effect |
| use dual-resolution (512 global, 1024 local) instead of single 1024 network (ablation) | GazeOnce360 · MPSGaze360 test split | FPS: 13.30 -> 16.23 FPS with gaze 8.945 -> 8.968 | Improved |
| use dual-resolution instead of single low-resolution (512) network (ablation) | GazeOnce360 · MPSGaze360 test split | gaze error (deg): 16.50 -> 8.968 | Improved |
| replace deformable convolution with rotational convolution (ablation) | GazeOnce360 · MPSGaze360 test split (D2) | gaze error (deg): 11.05 -> 10.39 | Improved |
| increase gaze loss weight from 1.0 to 10.0 (ablation) | GazeOnce360 · MPSGaze360 test split (D2) | gaze error (deg): 10.39 -> 7.741 | Improved |
| lower gaze loss weight from 1.0 to 0.5 (ablation) | GazeOnce360 · MPSGaze360 test split (D2) | gaze error (deg): 10.39 -> 11.15 | Hurt |
| change learning rate away from 1e-3 (ablation) | GazeOnce360 · MPSGaze360 test split (D2) | gaze error (deg): 10.39 -> 11.85 (5e-4); 13.16 (2e-3) | Hurt |
| evaluate on unseen scenes and identities instead of seen ones | GazeOnce360 · MPSGaze360 | gaze error (deg): 8.945 -> 10.39 | Hurt |
| replace multi-stage GAM360 pipeline with end-to-end GazeOnce360 | GazeOnce360 · MPSGaze360 D2 | gaze error (deg): 18.96 -> 10.39 | Improved |
Closest papers by meaning
- GA3CE: Unconstrained 3D Gaze Estimation with Gaze-Aware 3D Context Encoding (CVPR 2025 · similarity 0.87)
- Gaze-LLE: Gaze Target Estimation via Large-Scale Learned Encoders (CVPR 2025 · similarity 0.87)
- UVAGaze: Unsupervised 1-to-2 Views Adaptation for Gaze Estimation (AAAI 2024 · similarity 0.86)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (CVPR 2025 · similarity 0.86)
- MTGS: A Novel Framework for Multi-Person Temporal Gaze Following and Social Gaze Prediction (NeurIPS 2024 · similarity 0.86)
- Sharingan: A Transformer Architecture for Multi-Person Gaze Following (CVPR 2024 · similarity 0.85)
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.
