Research map / Video / Video segmentation and tracking
In the Eye of Transformer: Global–Local Correlation for Egocentric Gaze Estimation and Beyond
Bolin Lai, Miao Liu, Fiona Ryan, James Matthew Rehg
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 global visual token (strategy d, conv on local tokens) (ablation) | MViT · EGTEA Gaze+ | F1: 43.0 -> 43.9 | Improved |
| add global visual token (strategy d, conv on local tokens) (ablation) | MViT · Ego4D | F1: 40.9 -> 41.7 | Improved |
| use conv layers instead of max pooling for global token (ablation) | MViT + global token · EGTEA Gaze+ | F1: 43.4 -> 43.9 | Improved |
| add GLC module after encoder (ablation) | MViT + (d) · EGTEA Gaze+ | F1: 43.9 -> 44.8 | Improved |
| add GLC module after encoder (ablation) | MViT + (d) · Ego4D | F1: 41.7 -> 43.1 | Improved |
| add regular self-attention layer instead of GLC (parameter control) (ablation) | MViT + (d) · EGTEA Gaze+ | F1: 43.9 -> 44.1 | No clear effect |
| add regular self-attention layer instead of GLC (parameter control) (ablation) | MViT + (d) · Ego4D | F1: 41.7 -> 42.1 | No clear effect |
| replace SA layer with GLC (ablation) | MViT + (d) + SA · EGTEA Gaze+ | F1: 44.1 -> 44.8 | Improved |
| replace SA layer with GLC (ablation) | MViT + (d) + SA · Ego4D | F1: 42.1 -> 43.1 | Improved |
| add global token alone | MViT (cls token) · EGTEA Gaze+ action recognition | top-1 accuracy: 64.64 -> 64.44 | No clear effect |
| add global token alone | MViT (pooling) · EGTEA Gaze+ action recognition | top-1 accuracy: 63.45 -> 63.06 | Hurt |
| add GLC with pooling classifier (ablation) | MViT + Global Token · EGTEA Gaze+ action recognition | top-1 accuracy: 63.06 -> 65.33 | Improved |
Closest papers by meaning
- EgoM2P: Egocentric Multimodal Multitask Pretraining (ICCV 2025 · similarity 0.89)
- Ego-PMOVE: Prompt-aware Mixture of View Experts Network for Egocentric Gaze Prediction (AAAI 2026 · similarity 0.89)
- GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction (NeurIPS 2026 · similarity 0.88)
- Gaze Beyond the Frame: Forecasting Egocentric 3D Visual Span (NeurIPS 2025 · similarity 0.88)
- Gaze-VLM: Bridging Gaze and VLMs through Attention Regularization for Egocentric Understanding (NeurIPS 2025 · similarity 0.86)
- Object-Shot Enhanced Grounding Network for Egocentric Video (CVPR 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.
