extracted features of the indomain dataset VG

ChenRocks / UNITER

Research code for ECCV 2020 paper "UNITER: UNiversal Image-TExt Representation Learning"

https://arxiv.org/abs/1909.11740

779 stars 109 forks source link

Closed kellyiss closed 3 years ago

kellyiss commented 3 years ago

Hi, I find that you use dense captions with all the box features. Can I concate them?