Closed VV0808 closed 2 years ago
Thanks for your interest. I have to point out that each image is associated with 5 captions and each caption(maybe) has multiple relations. Because for flickr30k dataset, there are several noun phrases in one caption. Not all images have five captions but most cases do.
I have uploaded the train/val/test files. You can check it now
Thank you!
Hi! Thank you for your great work. I have some questions about sg_anno.json.
Each picture has five sentences.That means Each picture has five relations in sg_anno.json.
However, when i simply simply pair each image with their caption, like 1.jpg, 0; 1.jpg 1;1.jpg 2;1.jpg 3;1.jpg 4;
sent_id maybe not in sg_anno.json. there are only three or four relations in sg_anno.json, not five.
Even if there are less than five, are they one-to-one?
for example, there are five sentences about the picture. but there only three relations in sg_anno.json.
Do these three relationships correspond to the first three sentences?