861637a425ef06e6d539aaaff113d1d5-AuthorFeedback.pdf

We thank the reviewers for their comments. As agreed by all reviewers, it is an interesting and novel task, which will attract more researches to follow. Yes, we actually have adjusted DA to HGTL with a two-stage manner. All the compared methods and BigSPN are the two-stage models, as stated in Q2. Q1:Illustrating why a new visual encoder is needed.