Text-only synthesis for zero-shot visual captioning

Qing Zhou

doi:10.1016/j.patcog.2026.114985

Text-only synthesis for zero-shot visual captioning

2026-09-24

0

PRE

AI

DOI：10.1016/j.patcog.2026.114985

Original

Original request for help

Save

Abstract

En 中文

• ToCa requires fewer costs to generate large-scale training data for captioning models. • ToCa is a flexible and efficient algorithm for text synthesis. • ToCa demonstrates its impressive synthesis capabilities by addressing three application challenges.

Keywords:

Zero-shot

Visual captioning

Vision–language modeling

Text-only synthesis for zero-shot visual captioning

Abstract

Journal

Organization

Cited Papers