Decoding the Past: An Uncertainty-Aware Deep Learning Framework for Sex Attribution in Prehistoric Hand Stencils
arXiv: 2508.08543 Authors: Karel Becerra, Boris Mederos, Dean Snow Field: Computer Vision
Overview
Determining the biological sex of the individuals who created Upper Paleolithic hand stencils remains a challenging problem due to the absence of ground truth, population differences between contemporary and prehistoric groups, and the uncertainty introduced by image degradation. Traditional morphometric methods suffer from high structural overlap across sexes, poor cross-population generalizability, and subjective feature engineering.
This study presents an uncertainty-aware deep learning framework for sex attribution in prehistoric hand stencils that explicitly models, propagates, and aggregates uncertainty throughout the analytical pipeline.
Methodology
The approach combines:
- Dual image processing and dual contour extraction
- Structured silhouette augmentation to capture boundary uncertainty — 12 plausible contour realizations are generated per hand stencil
- Model architectural diversity: two ensembles of 10 deep neural networks each (EfficientNet-B3 and MobileViT-S), trained on 14,036 contemporary hand samples
- Triangulation scheme that integrates:
- Ensemble predictions
- Unsupervised 2D latent space manifold mapping (UMAP + k-NN)
- Explainable AI spatial attribution (LayerCAM) to ensure anatomical consistency
- On contemporary data, the ensemble models achieve strong classification performance, with accuracy exceeding 88% in the older age group.
- Applied to prehistoric hand stencils, the framework outputs both sex predictions and internal consistency confidence measures, distinguishing morphologically stable from ambiguous cases.
- The convergence of ensemble predictions, latent space structure, and explainability analysis shows that uncertainty can become a measurable component of archaeological inference, enabling robust and reproducible interpretation of ancient rock art.