Task-driven remote sensing image captioning for earth understanding: From multi-task dataset to task-conditioned framework

Z. Chen

doi:10.1016/j.isprsjprs.2026.05.024

Task-driven remote sensing image captioning for earth understanding: From multi-task dataset to task-conditioned framework

2026-05-27

0

PRE

AI

DOI：10.1016/j.isprsjprs.2026.05.024

Original

Original request for help

Save

Abstract

En 中文

• The multi-task captioning dataset is constructed with 1.06 million triplets across 13analytical tasks. • A task-conditioned captioning framework is proposed for multi-task urban analysis. • Cross-city and multi-task experiments demonstrate the effectiveness of the proposed dataset and method. • Applications on 20 global cities identified four patterns of city development stage types.

Keywords:

multi-task captioning

remote sensing image

urban analysis

task-conditioned framework

city development patterns

Task-driven remote sensing image captioning for earth understanding: From multi-task dataset to task-conditioned framework

Abstract

Journal

Organization