Polarization Human Shape and Pose Dataset

TL;DR

This paper introduces the Polarization Human Shape and Pose Dataset (PHSPD), a new dataset that explores the use of polarization images for detailed human shape and pose estimation, leveraging geometric cues from polarized reflected light.

Contribution

The paper presents a novel dataset that enables research on human shape estimation using polarization images, bridging polarization imaging and human modeling.

Findings

PHSPD provides diverse human shapes and poses for research.

Polarization cues can enhance detailed human shape reconstruction.

The dataset facilitates new methods leveraging polarization for human modeling.

Abstract

Polarization images are known to be able to capture polarized reflected lights that preserve rich geometric cues of an object, which has motivated its recent applications in reconstructing detailed surface normal of the objects of interest. Meanwhile, inspired by the recent breakthroughs in human shape estimation from a single color image, we attempt to investigate the new question of whether the geometric cues from polarization camera could be leveraged in estimating detailed human body shapes. This has led to the curation of Polarization Human Shape and Pose Dataset (PHSPD), our home-grown polarization image dataset of various human shapes and poses.