Research map / Detection and segmentation / Open-set and few-shot detection
Gazeformer: Scalable, Effective and Fast Prediction of Goal-Directed Human Attention
Sounak Mondal, Zhibo Yang, Seoyoung Ahn, Dimitris Samaras, Gregory Zelinsky, Minh Hoai
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 |
|---|---|---|---|
| Use RoBERTa target embeddings instead of fixed random embeddings (ablation) | Gazeformer · COCO-Search18 ZeroGaze | SS: 0.336 -> 0.359 | Improved |
| Use RoBERTa target embeddings instead of fixed random embeddings (ablation) | Gazeformer · COCO-Search18 ZeroGaze | FED: 2.873 -> 2.788 | Improved |
| Use RoBERTa target embeddings instead of fixed random embeddings (ablation) | Gazeformer · COCO-Search18 ZeroGaze | NSS: 4.524 -> 4.671 | Improved |
| Replace patch classification with coordinate regression (ablation) | Gazeformer · COCO-Search18 GazeTrain | SS: 0.477 -> 0.504 | Improved |
| Replace patch classification with coordinate regression (ablation) | Gazeformer · COCO-Search18 GazeTrain | FED: 2.158 -> 2.072 | Improved |
| Replace patch classification with coordinate regression (ablation) | Gazeformer · COCO-Search18 GazeTrain | NSS: 7.545 -> 8.375 | Improved |
| Do not model fixation duration (ablation) | Gazeformer · COCO-Search18 ZeroGaze | SS (w/o Dur): 0.358 -> 0.359 | No clear effect |
| Do not model fixation duration (ablation) | Gazeformer · COCO-Search18 GazeTrain | SS (w/o Dur): 0.504 -> 0.504 | No clear effect |
| Use parallel transformer decoding to generate the scanpath | Gazeformer · COCO-Search18 target-present test split | inference time (ms): 386 -> 68 ms; 5.7X | Improved |
| Use RoBERTa semantic target features to predict unseen target categories | Gazeformer · COCO-Search18 ZeroGaze | SS: 0.300 -> 0.358 | Improved |
Later papers that compared with this method
1 later accepted paper reports Gazeformer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- OAT: Object-Level Attention Transformer for Gaze Scanpath Prediction (ECCV 2024 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning from Observer Gaze: Zero-Shot Attention Prediction Oriented by Human-Object Interaction Recognition (CVPR 2024 · similarity 0.91)
- Unifying Top-down and Bottom-up Scanpath Prediction Using Transformers (CVPR 2024 · similarity 0.90)
- MTGS: A Novel Framework for Multi-Person Temporal Gaze Following and Social Gaze Prediction (NeurIPS 2024 · similarity 0.90)
- EyEar: Learning Audio Synchronized Human Gaze Trajectory Based on Physics-Informed Dynamics (AAAI 2025 · similarity 0.89)
- Gaze-LLE: Gaze Target Estimation via Large-Scale Learned Encoders (CVPR 2025 · similarity 0.89)
- Sharingan: A Transformer Architecture for Multi-Person Gaze Following (CVPR 2024 · similarity 0.89)
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.
