Enhancing Close-up Novel View Synthesis via Pseudo-labeling

Jiatong Xia, Libo Sun, Lingqiao Liu · 2025

Read Enhancing Close-up Novel View Synthesis via Pseudo-labeling by Jiatong Xia, Libo Sun, Lingqiao Liu (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-2503-15908.txt.

Description

Jiatong Xia, Libo Sun, and Lingqiao Liu improve a weak spot in 3D image making. Methods such as Neural Radiance Fields and 3D Gaussian Splatting build scenes that can be viewed from new angles. Close-up views, though, often come out blurred or wrong. Training photos rarely include true close shots.

Their fix teaches the model with pseudo-labels, extra training targets drawn from the photos it already has, aimed at a wide range of close viewpoints. Since no test set existed for this task, they also built a new benchmark data set for close-up view quality. In their tests, the approach sharpened close views well beyond the base methods. The work matters for uses such as product viewing, virtual tours, and any scene where users lean in for detail.

Catalog record

CREATORJiatong Xia, Libo Sun, Lingqiao Liu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-15908

Browse the full library · About the Tower of Babel