Multipolar

The MultiPolar dataset is a specialized collection of optical data designed to facilitate the development and validation of machine learning models within the field of experimental mechanics. Its primary purpose is to bridge the gap between traditional Digital Photoelasticity and modern data-driven frameworks by providing a standardized set of multi-polarizer imagery and their corresponding quantitative stress interpretations.

This dataset includes multi-polarizer array images consisting of raw data captured from optical setups, featuring both standard benchmark geometries and intricate bioinspired specimens. These samples cover a wide range of stress concentration scenarios, represented through diverse fringe pattern images to ensure model robustness across different structural complexities.

While Digital Photoelasticity has transitioned toward using multipolarizer array cameras, the lack of open-access training data has slowed the integration of Artificial Intelligence in stress analysis. The MultiPolar Dataset addresses this by offering a foundation for: (a) Automated demodulation, by training models to convert fringe patterns into stress maps; (b) Benchmark testing, by providing a common ground for comparing new algorithms in experimental mechanics; and (c) Reproducibility, by promoting open-science practices in the study of material response under stress.