Is Geometry Enough? An Evaluation of Landmark-Based Gaze Estimation

arXiv cs.AI / 3/27/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper evaluates landmark-based (geometric) gaze estimation as a lightweight, more interpretable alternative to appearance-based CNN approaches, which are accurate but computationally expensive and less transparent.
It proposes a standardized landmark extraction and normalization pipeline across three large datasets (Gaze360, ETH-XGaze, and GazeGene) and trains lightweight regressors (XGBoost) alongside two neural landmark-based models (holistic MLP and a binocular-geometry siamese MLP).
Results show landmark-based methods underperform in within-domain tests, plausibly because landmark detector noise degrades training and evaluation data.
In cross-domain experiments, the proposed landmark-based MLPs generalize similarly to ResNet18 baselines, indicating that sparse geometric features can carry enough signal for robust gaze estimation.
The authors release code and generated landmark datasets, arguing the approach can support efficient, interpretable, and privacy-friendly edge deployments.

Abstract

Appearance-based gaze estimation frequently relies on deep Convolutional Neural Networks (CNNs). These models are accurate, but computationally expensive and act as "black boxes", offering little interpretability. Geometric methods based on facial landmarks are a lightweight alternative, but their performance limits and generalization capabilities remain underexplored in modern benchmarks. In this study, we conduct a comprehensive evaluation of landmark-based gaze estimation. We introduce a standardized pipeline to extract and normalize landmarks from three large-scale datasets (Gaze360, ETH-XGaze, and GazeGene) and train lightweight regression models, specifically Extreme Gradient Boosted trees and two neural architectures: a holistic Multi-Layer Perceptron (MLP) and a siamese MLP designed to capture binocular geometry. We find that landmark-based models exhibit lower performance in within-domain evaluation, likely due to noise introduced into the datasets by the landmark detector. Nevertheless, in cross-domain evaluation, the proposed MLP architectures show generalization capabilities comparable to those of ResNet18 baselines. These findings suggest that sparse geometric features encode sufficient information for robust gaze estimation, paving the way for efficient, interpretable, and privacy-friendly edge applications. The source code and generated landmark-based datasets are available at https://github.com/daniele-agostinelli/LandmarkGaze.git.