What is the ground truth when I use natural language queries to retrieve the source image？

In your paper, your dense captioning model can support image retrieval using natural language queries, and can localize these queries in retrieved images. What the ground truth when you calculate R@n?

Start by reading the paper's description of natural-language image retrieval and its R@n evaluation. Determine which retrieved image or region is treated as ground truth, then document that definition and how the metric is calculated; the payload names no repository file or test to update.

Written by the indexing model from the issue text.