amitha/docent-eval-coarse
收藏官方服务：
资源简介：
Docent Eval是一个用于评估详细图像描述度量的基准，它包含了对艺术品详细描述的粗粒度判断。该数据集用于确保详细图像描述度量是评价人类评分的良好代理。数据集中的每一行对应于一幅艺术品的一对生成的详细描述的粗粒度评价，评价是基于与专家编写的参考描述的比较。
Docent Eval: a benchmark for evaluating detailed image description metrics which contain coarse judgments of detailed descriptions for works of art. It is used to ensure that detailed image description metrics are a good proxy for human raters. Each row in the dataset corresponds to a coarse judgment of a pair of generated detailed descriptions for a work of art, based on comparison with an expert-written reference description.
提供机构：
amitha
