BriefGPT.xyz
Oct, 2016
不透过物体生成标题
Generating captions without looking beyond objects
HTML
PDF
Hendrik Heuer, Christof Monz, Arnold W. M. Smeulders
TL;DR
该论文探讨图像字幕的新评估视角,并引入了一项名词翻译任务,通过从一组名词到字幕的翻译实现了相对图像字幕生成性能,还研究了称作 BLEU 的指标中各个单词类别的下限和上限。
Abstract
This paper explores new
evaluation
perspectives for
image captioning
and introduces a
noun translation
task that achieves comparative imag
→