Reproduce “NeutrEx: A 3D Quality Component Measure on Facial Expression Neutrality” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Marcel Grimmer, Christian Rathgeb, Raymond Veldhuis, Christoph Busch
Grimmer et al. introduce NeutrEx, a 3D-based quality measure designed to quantify facial expression neutrality for face recognition systems, aligning with ISO/IEC 29794-5 standards. The method consistently outperforms 2D baseline approaches in predicting face recognition utility and offers explainable feedback through per-vertex visualizations of facial deformation.
Key insights
- Quantifying facial expression neutrality through 3D geometric deviation provides a more precise prediction of face recognition utility compared to 2D embedding-based confidence scores.
- The full-vertex correspondence in 3D face models allows for explainable quality measures, visualizing specific facial regions most impacted by non-neutral expressions.
- While 2D baselines can classify expressions, their limited ability to capture subtle geometric changes from neutrality results in less effective prediction of FR system performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.