UnrealPose: Leveraging Game Engine Kinematics for Large-Scale Synthetic Human Pose Data

2026

Read UnrealPose: Leveraging Game Engine Kinematics for Large-Scale Synthetic Human Pose Data (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-00991.txt.

Description

Joshua Kawaguchi and seven colleagues manufacture human pose data inside a video game engine instead of a motion capture studio. Labeled three-dimensional pose data are expensive and studio-bound, while outdoor datasets lack exact ground truth. Their pipeline, UnrealPose-Gen, renders frames with Unreal Engine 5 and records everything a pose model needs to learn.

Each generated frame carries three-dimensional joints in world and camera coordinates, two-dimensional keypoints with occlusion flags, bounding boxes, and the camera's full parameters. With it they built UnrealPose-1M, a corpus of roughly one million frames spanning scripted coherent sequences and randomized ones, several scenes, dozens of actions, and five subjects, shot from varied camera paths. A fidelity check compares the synthetic data against real captures.

Catalog record

YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-00991

Browse the full library · About the Tower of Babel