Research map / Vision-language and multimodal / CLIP and zero-shot
LG-Gaze: Learning Geometry-aware Continuous Prompts for Language-Guided Gaze Estimation
Pengwei Yin, Jingjing Wang, Guanzhong Zeng, Di Xie, Jiang Zhu
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 MCR loss with language guidance to the gaze loss (ablation) | ResNet-18 gaze model · Four cross-domain tasks | average angular error (deg): 8.65 -> 7.01 | Improved |
| add geometry loss on anchor embeddings (ablation) | LG-Gaze · Four cross-domain tasks | average angular error (deg): 7.01 -> 6.84 | Improved |
| use geometry-aware spherical interpolation instead of global linear interpolation (ablation) | LG-Gaze · Four cross-domain tasks | average angular error (deg): 7.34 -> 6.84 | Improved |
| use spherical interpolation instead of bilinear interpolation (ablation) | LG-Gaze · Four cross-domain tasks | average angular error (deg): 6.98 -> 6.84 | Improved |
| add 256 global negative samples (ablation) | MCR loss · Four cross-domain tasks | average angular error (deg): 7.21 -> 6.84 | Improved |
| increase global negatives from 256 to 1024 (ablation) | MCR loss · Four cross-domain tasks | average angular error (deg): 6.84 -> 6.81 | No clear effect |
| use MCR instead of CDG, RNC or MoCo contrastive losses (ablation) | LG-Gaze · Four cross-domain tasks | average angular error (deg): 6.96 -> 6.84 | Improved |
| apply CoOp text prompt tuning to gaze | CLIP · Four cross-domain tasks | average angular error (deg): 8.65 -> 11.07 | Hurt |
| use zero-shot Vanilla CLIP for gaze prediction | CLIP · Four cross-domain tasks | average angular error (deg): 8.65 -> 81.04 | Hurt |
| use LG-Gaze versus CLIP-Gaze on ETH-XGaze->MPIIFaceGaze | LG-Gaze · ETH-XGaze -> MPIIFaceGaze | angular error (deg): 6.41 -> 6.45 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CLIP-Gaze — CLIP-Gaze: Towards General Gaze Estimation via Visual-Linguistic Model (AAAI 2024) · better on 1 of 1 shared result
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
- CLIP-Gaze: Towards General Gaze Estimation via Visual-Linguistic Model (AAAI 2024 · similarity 0.91)
- Test-Time Personalization with Meta Prompt for Gaze Estimation (AAAI 2024 · similarity 0.88)
- Gaze from Origin: Learning for Generalized Gaze Estimation by Embedding the Gaze Frontalization Process (AAAI 2024 · similarity 0.88)
- From Feature to Gaze: A Generalizable Replacement of Linear Layer for Gaze Estimation (CVPR 2024 · similarity 0.88)
- OmniGF: A Dual-Branch Vision-Language Framework for Unified Gaze Following (NeurIPS 2026 · similarity 0.87)
- Eyes on VLM: Benchmarking Gaze Following and Social Gaze Prediction in Vision Language Models (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.
