2DGS-Avatar: Animatable High-fidelity Clothed Avatar via 2D Gaussian Splatting

Qipeng Yan, Mingyang Sun, Lihua Zhang* · 2025

Read 2DGS-Avatar: Animatable High-fidelity Clothed Avatar via 2D Gaussian Splatting by Qipeng Yan, Mingyang Sun, Lihua Zhang* (2025) 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-2503-02452.txt.

Description

Qipeng Yan, Mingyang Sun, and Lihua Zhang build animatable avatars from phone-style video in a 2025 paper. Rendering a clothed person who can be re-posed in real time is a long-standing graphics goal. Neural field methods look good but render slowly. Splatting methods, which paint scenes from thousands of soft blobs, run fast yet can warp the shape.

Their method uses flat, two-dimensional splats fitted to the body, learning from a single camera's video. The avatar then follows new poses while keeping detailed, photo-like clothing. Tests on two popular avatar datasets showed strong scores on both looks and numbers. Fast training rounds out the appeal for creators who need a digital double without a studio.

Catalog record

CREATORQipeng Yan, Mingyang Sun, Lihua Zhang*
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-02452

Browse the full library · About the Tower of Babel