AI Paper Atlas

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

IJCV 2023 · Video › Video segmentation and tracking · approach: Learned optical flow estimation · 43 citations (OpenAlex)

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 changedWhereMeasuredEffect
add global visual token (strategy d, conv on local tokens) (ablation)MViT · EGTEA Gaze+F1: 43.0 -> 43.9Improved
add global visual token (strategy d, conv on local tokens) (ablation)MViT · Ego4DF1: 40.9 -> 41.7Improved
use conv layers instead of max pooling for global token (ablation)MViT + global token · EGTEA Gaze+F1: 43.4 -> 43.9Improved
add GLC module after encoder (ablation)MViT + (d) · EGTEA Gaze+F1: 43.9 -> 44.8Improved
add GLC module after encoder (ablation)MViT + (d) · Ego4DF1: 41.7 -> 43.1Improved
add regular self-attention layer instead of GLC (parameter control) (ablation)MViT + (d) · EGTEA Gaze+F1: 43.9 -> 44.1No clear effect
add regular self-attention layer instead of GLC (parameter control) (ablation)MViT + (d) · Ego4DF1: 41.7 -> 42.1No clear effect
replace SA layer with GLC (ablation)MViT + (d) + SA · EGTEA Gaze+F1: 44.1 -> 44.8Improved
replace SA layer with GLC (ablation)MViT + (d) + SA · Ego4DF1: 42.1 -> 43.1Improved
add global token aloneMViT (cls token) · EGTEA Gaze+ action recognitiontop-1 accuracy: 64.64 -> 64.44No clear effect
add global token aloneMViT (pooling) · EGTEA Gaze+ action recognitiontop-1 accuracy: 63.45 -> 63.06Hurt
add GLC with pooling classifier (ablation)MViT + Global Token · EGTEA Gaze+ action recognitiontop-1 accuracy: 63.06 -> 65.33Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy