Research map / 3D vision / Human pose estimation
Following Gaze in Video
Adria Recasens, Carl Vondrick, Aditya Khosla, Antonio Torralba
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 the saliency pathway to cone and transformation pathways (ablation) | Gaze network · VideoGaze test (gaze location) | AUC: 83.8 -> 89.0 | Improved |
| add the cone and transformation pathways to the saliency pathway (ablation) | Gaze network · VideoGaze test (gaze location) | AUC: 87.7 -> 89.0 | Improved |
| add a learned affine transformation instead of identity (ablation) | Gaze network · VideoGaze test (gaze location) | AUC: 86.6 -> 89.0 | Improved |
| use full 3-axis rotation plus translation instead of translation only (ablation) | Gaze network · VideoGaze test (gaze location) | AUC: 87.9 -> 89.0 | Improved |
| add translation to 3-axis rotation (ablation) | Gaze network · VideoGaze test (gaze location) | AUC: 87.5 -> 89.0 | Improved |
| use 3-axis rotation instead of vertical-axis rotation only (ablation) | Gaze network · VideoGaze test (gaze location) | AUC: 88.5 -> 89.0 | No clear effect |
| add saliency pathway for frame selection (ablation) | Frame selector · VideoGaze test (frame selection) | AP: 86.7 -> 87.5 | Improved |
| use source frame as well as target frame for frame selection (ablation) | Frame selector · VideoGaze test (frame selection) | AP: 83.9 -> 87.5 | Improved |
| use 3-axis rotation instead of vertical-axis rotation only for frame selection (ablation) | Frame selector · VideoGaze test (frame selection) | AP: 87.1 -> 87.5 | No clear effect |
| add saliency pathway to cone pathway (ablation) | Gaze network · VideoGaze test (gaze location) | L2 distance: 0.194 -> 0.184 | Improved |
| add cone and transformation pathways to saliency pathway (ablation) | Gaze network · VideoGaze test (gaze location) | L2 distance: 0.236 -> 0.184 | Improved |
Closest papers by meaning
- MTGS: A Novel Framework for Multi-Person Temporal Gaze Following and Social Gaze Prediction (NeurIPS 2024 · similarity 0.90)
- Toward Semantic Gaze Target Detection (NeurIPS 2024 · similarity 0.89)
- Sharingan: A Transformer Architecture for Multi-Person Gaze Following (CVPR 2024 · similarity 0.88)
- GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction (NeurIPS 2026 · similarity 0.88)
- Eyes on VLM: Benchmarking Gaze Following and Social Gaze Prediction in Vision Language Models (NeurIPS 2026 · similarity 0.87)
- ST-DiffEye: Diffusion-based Continuous Gaze Generation via Joint Scanpath-Trajectory Modeling (NeurIPS 2026 · similarity 0.87)
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.
