Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Gaze Estimation via Disentangled Subspace Contrastive Learning
Qida Tan, Hongyu Yang, Wenchao Du
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 DSCL to supervised baseline | ResNet-18 baseline · Gaze360, 5% labels | MAE (degrees): 23.00 -> 14.21 | Improved |
| add DSCL to existing gaze methods | Gazetr, RT-Gene, FullFace, Dilated-Net, ST-WSGE · Gaze360 (20/10/5% labels) | MAE (degrees): avg 16.57 -> 15.14 (Gazetr), 17.08 -> 14.79 (ST-WSGE) | Improved |
| add DSCL to cross-domain baselines | Baseline DG, PureGaze, PnP-GA · Gaze360 -> MPIIGaze / EyeDiap | MAE (degrees): 18.75 -> 10.21 (baseline, EyeDiap 5%) | Improved |
| remove Jacobian regularization LJ (ablation) | DSCL · Gaze360, 5% labels | MAE (degrees): 17.66 -> 14.21 | Improved |
| add unsupervised ranking loss LUR (ablation) | DSCL · Gaze360, 5% labels | MAE (degrees): 18.88 -> 14.21 | Improved |
| add unsupervised contrastive loss LUC (ablation) | DSCL · Gaze360, 5% labels | MAE (degrees): 15.97 -> 14.21 | Improved |
| add supervised contrastive loss LSC (ablation) | DSCL · Gaze360, 5% labels | MAE (degrees): 15.40 -> 14.21 | Improved |
| add both unlabeled-data losses (LUR and LUC) (ablation) | DSCL · Gaze360, 5% labels | MAE (degrees): 21.72 -> 14.21 | Improved |
| add supervised initialization phase (Lreg + LJ) (ablation) | DSCL · Gaze360, 10% labels | MAE (degrees): 19.96 -> 13.94 | Improved |
| use DSCL instead of other semi-supervised methods | ResNet-18 baseline · Gaze360, 5% labels | MAE (degrees): CLSS 20.77 -> DSCL 14.21; UCVME 21.49; RankUp 24.95 | Improved |
| add external unlabeled data (WebFace and CelebA) (ablation) | DSCL · Gaze360 -> MPIIGaze, 10% labels | MAE (degrees): 10.57 -> 9.15 | Improved |
| add WebFace only as external unlabeled data (ablation) | DSCL · Gaze360, 5% labels | MAE (degrees): 14.21 -> 14.26 | 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.
- RankUp — RankUp: Boosting Semi-Supervised Regression with an Auxiliary Ranking Classifier (NeurIPS 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
- Bootstrap AutoEncoders With Contrastive Paradigm for Self-supervised Gaze Estimation (ICML 2024 · similarity 0.87)
- Gaze from Origin: Learning for Generalized Gaze Estimation by Embedding the Gaze Frontalization Process (AAAI 2024 · similarity 0.87)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (CVPR 2025 · similarity 0.87)
- OmniGaze: Reward-inspired Generalizable Gaze Estimation in the Wild (NeurIPS 2025 · similarity 0.87)
- From Feature to Gaze: A Generalizable Replacement of Linear Layer for Gaze Estimation (CVPR 2024 · similarity 0.86)
- De^2Gaze: Deformable and Decoupled Representation Learning for 3D Gaze Estimation (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.
