Search papers, labs, and topics across Lattice.
This paper introduces a curvature-aware spectral framework for evaluating 3D Morphable Models (3DMMs) in face reconstruction, addressing the inherent morphological biases that arise from limited training data. By utilizing the Laplace-Beltrami Operator (LBO), the authors generate high-resolution curvature error maps that reveal subtle discrepancies in surface topology, leading to a more accurate assessment of reconstruction quality. The validation through user studies shows that this new error metric correlates significantly better with human perception than traditional Euclidean distance measures, uncovering systematic biases related to age, gender, and ethnicity in 3D face models.
Curvature-aware evaluation reveals hidden biases in 3D face reconstruction, challenging the fairness of existing models across diverse populations.
3D Morphable Models (3DMMs) remain the standard parametric shape priors for many state-of-the-art 3D face reconstruction algorithms. However, as these models are derived from a finite number of 3D face samples, they inherit the morphological biases of their training data, potentially limiting their generalizability across diverse global populations. In this paper, we propose a novel framework to analyze 3DMM reconstructions through the lens of surface curvature, with the objective to discover, quantify and visualize biases. While standard evaluation metrics often rely on Euclidean distances, our reconstruction error captures subtle surface nuances such as local topology or undulations. To do so, we leverage the Laplace-Beltrami Operator (LBO) to generate high-resolution curvature error maps, providing a localized and geometrically meaningful visualization of discrepancies between ground truth faces and reconstructed meshes. We derive from it an error metric that we validated through a user study, observing a significantly higher correlation to human perception compared to traditional methods. Furthermore, we conduct extensive experiments across several 3DMM bases and fitting algorithms, uncovering systematic age-related biases and providing preliminary evidence of biases associated with gender and ethnicity. Our findings highlight the necessity of adopting curvature-aware evaluation protocols to ensure demographic fairness and geometric precision in future 3D face reconstruction research.
