Research map / 3D vision / Depth and stereo
Look Both Ways: Self-Supervising Driver Gaze Estimation and Road Scene Saliency
Isaac Kasahara, Simon Stent, Hyun Soo Park
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 geometric gaze-saliency self-supervision | supervised-only ETH XGaze · LBW gaze, self-supervised split (5% supervised) | gaze MAE (degrees): 9.7 -> 8.2 | Improved |
| add geometric self-supervision | supervised-only ETH XGaze · LBW gaze, self-supervised split (60% supervised) | gaze MAE (degrees): 7.0 -> 6.2 | Improved |
| add geometric self-supervision | supervised-only ETH XGaze · LBW gaze, self-supervised split (40% supervised) | gaze MAE (degrees): 8.1 -> 7.4 | Improved |
| add geometric self-supervision | supervised-only ETH XGaze · LBW gaze, self-supervised split (20% supervised) | gaze MAE (degrees): 6.9 -> 6.9 | No clear effect |
| add geometric self-supervision | supervised-only ETH XGaze · LBW gaze, test split (5% supervised) | gaze MAE (degrees): 7.8 -> 7.8 | No clear effect |
| add geometric self-supervision | supervised-only ETH XGaze · LBW gaze, test split (20% supervised) | gaze MAE (degrees): 6.8 -> 6.5 | Improved |
| add geometric self-supervision | supervised-only ETH XGaze · LBW gaze, test split (60% supervised) | gaze MAE (degrees): 6.7 -> 6.7 | No clear effect |
| add geometric self-supervision | supervised-only Unisal · LBW saliency, test split (60% supervised) | saliency MAE: 1.03 -> 1.03 | No clear effect |
| add geometric self-supervision | supervised-only Unisal · LBW saliency, self-supervised split (5% supervised) | saliency MAE: 1.14 vs 1.16 on test; 1.12 vs 1.14 on self-supervised split | Improved |
| fine-tune on in-domain labeled LBW data | ETH XGaze · LBW gaze, test split (5% supervised) | gaze MAE (degrees): 15.6 -> 7.8 | Improved |
| fine-tune on in-domain labeled LBW data | Unisal · LBW saliency, test split (5% supervised) | saliency MAE: 1.60 -> 1.16 | Improved |
| use supervised-only gaze model instead of mean-gaze predictor | ETH XGaze · LBW gaze, test split (5% supervised) | gaze MAE (degrees): 9.2 -> 7.8 | Improved |
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.
- Gaze360 — Gaze360: Physically Unconstrained Gaze Estimation in the Wild (ICCV 2019) · better on 4 of 4 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
- "Looking at the Right Stuff" - Guided Semantic-Gaze for Autonomous Driving (CVPR 2020 · similarity 0.91)
- FIFA: Fine-grained Inter-frame Attention for Driver's Video Gaze Estimation (CVPR 2025 · similarity 0.90)
- What Do You See in Vehicle? Comprehensive Vision Solution for In-Vehicle Gaze Estimation (CVPR 2024 · similarity 0.90)
- Leveraging Driver Field-of-View for Multimodal Ego-Trajectory Prediction (ICLR 2025 · similarity 0.87)
- On-Road Object Importance Estimation: A New Dataset and A Model with Multi-Fold Top-Down Guidance (NeurIPS 2024 · similarity 0.86)
- Rethinking Lanes and Points in Complex Scenarios for Monocular 3D Lane Detection (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.
