format explanation for the grounding annotation file
Closed this issue · 3 comments
peiwang062 commented
any explanation or instruction for how to understand the format of the annotation files, expression
and labels_with_ids
?
how to link each other.
wudongming97 commented
Each expression corresponds to multiple frames, each frame containing multiple labels_with_ids.
Sorry for the confusion, and we will update the details.
peiwang062 commented
Thanks for the reply!
I still have no idea where to start. For example, for an expression, black-cars-in-right.json, how can I find its corresponding image and associated bbox?
wudongming97 commented
When you open a JSON file, its corresponding image sequence names can be found in terms of its folder name.
Our JSON file have only corresponding ids, so the corresponding box can be seen from 'label_with_ids'.