FigEx2: Visual-Conditioned Panel Detection and Captioning for Scientific Compound Figures

Jifeng Song, Arun Das, Pan Wang, Hui Ji, Kun Zhao, Yufei Huang · 2026

Read FigEx2: Visual-Conditioned Panel Detection and Captioning for Scientific Compound Figures by Jifeng Song, Arun Das, Pan Wang, Hui Ji, Kun Zhao, Yufei Huang (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-2601-08026.txt.

Description

Jifeng Song and colleagues present FigEx2, a system that finds and captions each panel of a scientific compound figure. Compound figures pack several labeled panels into one image. In real pipelines, captions are often missing or only summarize the whole figure. Panel-level understanding is therefore difficult. FigEx2 localizes panels and generates a caption for each one, directly from the figure image. Their 2026 preprint describes the visual-conditioned framework.

Panel captions unlock the content of published figures for search and reading tools. Accessibility also gains, since screen readers need panel-level descriptions. Conditioning on the visual layout keeps captions tied to the right panel. The work targets the messy figures found in real journals. The paper serves scholars in document AI and scientific publishing tools. It treats figures as structured content, not decoration.

Catalog record

CREATORJifeng Song, Arun Das, Pan Wang, Hui Ji, Kun Zhao, Yufei Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-08026

Browse the full library · About the Tower of Babel