Densepol dataset boosts accuracy of polarization from rgb images
DensePol: Dense-Angle Polarization Dataset for Learning-Based Polarimetric Vision
Computer Vision and Pattern Recognition
Summary
Polarization in vision helps reveal shapes and materials that regular color images cannot easily show. The authors created a new dataset called DensePol that captures detailed polarization information by measuring many angles carefully. This dataset reduces errors in polarization data, helping computer programs learn to predict polarization more accurately from ordinary RGB images. They also developed a new method that uses this dataset to improve predictions and estimate surface shapes better.
What this means in practice
- •For robotics developers: Enhance robot perception by accurately estimating surface shapes and materials from standard RGB cameras using DensePol-trained models.
- •For augmented reality engineers: Improve scene understanding and realistic rendering by predicting polarization cues from regular RGB data in real time with DensePol-based methods.
Authors
Param Sangani, Ahmad Moori, Erik Blasch, Guna Seetharaman, Hadi Aliakbarpour
Abstract
Polarimetric vision is gaining increasing attention because it provides physical cues about scene shape, material, and reflection that are difficult to recover from RGB alone. Recent work has therefore explored predicting polarization directly from conventional RGB images; however, the fidelity of these methods strongly depends on the polarization supervision used for training. Most existing datasets rely on Division-of-Focal-Plane (DoFP) cameras with four spatially interleaved analyzer orientations, which provide limited angular redundancy and introduce interpolation and instantaneous-field-of-view errors. We introduce DensePol, a high-redundancy RGB--polarization dataset based on Division-of-Time (DoT) acquisition, capturing 180 full-resolution analyzer orientations at $1^\circ$ intervals. DensePol contains 2,018 paired RGB--polarization images with the angular measurements and fitting residuals retained. Dense angular sampling substantially improves polarization stability, reducing AoLP deviation from $13.36^\circ$ to $2.21^\circ$. We further introduce a deterministic diffusion-based RGB-to-polarization framework with cyclic AoLP representation and a local DoLP refiner. Experiments demonstrate improved polarization prediction and downstream surface-normal estimation. The dataset and code will be publicly available.