Research map / Video / Video segmentation and tracking
Predicting Gaze in Egocentric Video by Learning Task-dependent Attention Transition
Yifei Huang, Minjie Cai, Zhenqiang Li, Yoichi Sato
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 attention transition module to saliency prediction (full model) (ablation) | SP · GTEA Gaze Plus | AAE (deg): 5.52 -> 4.05 | Improved |
| add attention transition module to saliency prediction (full model) (ablation) | SP · GTEA Gaze Plus | AUC: 0.928 -> 0.957 | Improved |
| add attention transition module to saliency prediction (full model) (ablation) | SP · GTEA Gaze | AAE (deg): 9.43 -> 7.58 | Improved |
| add attention transition module to saliency prediction (full model) (ablation) | SP · GTEA Gaze | AUC: 0.861 -> 0.898 | Improved |
| add saliency prediction module to attention transition (full model) (ablation) | AT · GTEA Gaze Plus | AAE (deg): 5.02 -> 4.05 | Improved |
| use modified bce loss instead of plain bce for two-stream saliency (ablation) | SP · GTEA Gaze Plus | AUC: 0.918 -> 0.928 | Improved |
| use modified bce loss instead of plain bce for spatial stream (ablation) | S-CNN · GTEA Gaze Plus | AUC: 0.893 -> 0.905 | Improved |
| use modified bce loss instead of plain bce for temporal stream (ablation) | T-CNN · GTEA Gaze Plus | AAE (deg): 6.15 -> 6.07 | Improved |
| use modified bce loss instead of plain bce (ablation) | T-CNN · GTEA Gaze Plus | AUC: 0.906 -> 0.906 | No clear effect |
| combine spatial and temporal streams (SP) versus spatial stream alone (ablation) | S-CNN · GTEA Gaze Plus | AUC: 0.905 -> 0.928 | Improved |
| attention transition module alone versus saliency prediction alone (ablation) | SP · GTEA Gaze | AAE (deg): 9.43 -> 9.51 | No clear effect |
| train attention transition on the same task rather than different tasks (ablation) | AT · GTEA Gaze Plus cross-task | AAE/AUC | Improved |
Closest papers by meaning
- Ego-PMOVE: Prompt-aware Mixture of View Experts Network for Egocentric Gaze Prediction (AAAI 2026 · similarity 0.90)
- GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction (NeurIPS 2026 · similarity 0.89)
- Gaze-VLM: Bridging Gaze and VLMs through Attention Regularization for Egocentric Understanding (NeurIPS 2025 · similarity 0.88)
- Gaze Beyond the Frame: Forecasting Egocentric 3D Visual Span (NeurIPS 2025 · similarity 0.87)
- In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting (NeurIPS 2025 · similarity 0.86)
- DiffEye: Diffusion-Based Continuous Eye-Tracking Data Generation Conditioned on Natural Images (NeurIPS 2025 · 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.
